Patterns of CBRN Use by Non-State Actors:

Size: px
Start display at page:

Download "Patterns of CBRN Use by Non-State Actors:"

Transcription

1 ISA 2013 Patterns of CBRN Use by Non-State Actors: Analyzing the Evidence Ronald Breiger 1 Paul Murray 1 Lauren Pinson 2 1 University of Arizona, Tucson, AZ National Consortium for the Study of Terrorism and Responses to Terrorism (START), University of Maryland, College Park, MD Paper prepared for presentation at the Annual Convention of the International Studies Association, San Francisco, April 4, 2013, Panel on New Data for the Scientific Study of Conflict This work was supported by the Defense Threat Reduction Agency, Basic Research Award # HDTRA We thank Gary Ackerman, Victor Asal, David Melamed, H. Brinton Milward, Karl Rethemeyer, and Eric Schoon for helpful discussions.

2 Patterns of CBRN Use by Non-State Actors: Analyzing the Evidence In recent years, academics and policymakers have debated the growing potential for chemical, biological, radiological, and nuclear (CBRN) terrorist attacks. However, CBRN terrorism is often misrepresented by the media due to limited first-hand information, conflicting reports, and varying sympathies which creates unreliable events data. This paper has two goals. The first is to introduce the reader to a new database for the scientific study of activities (plots, acquisitions, weaponization, attacks) of non-state actors seeking or using CBRN agents. Analysts have often attempted to mine CBRN databases and related datasets on terrorism by making use of standard regression models and their many generalizations such as logistic regression for binary outcomes (e.g., Asal and Rethemeyer 2009, on CBRN use or pursuit by Islamist organizations), Poison regression for counts (e.g., Asal & Rethemeyer 2008, modeling attack lethality), and many more. Our second goal therefore is to illustrate a set of newly developed techniques for, so to speak, turning these regression models inside out so that instead of focusing on relations among variables the analyst can use those relations to model a network of profile similarity among the cases (which for us are specific events). In this respect we join with a number of other papers presented in this panel that seek to extend the conventional quantitative modeling techniques in a variety of new directions. Philip Schrodt (2012, p. 556) has recently noted that event data is perfectly suited for network analysis. We intend to introduce an implementation of this idea that is quite non-standard, but one which we have found to be useful in working with databases in analyzing conflict situations. In illustrating the modeling of a new events databases, we build on our previous modeling work (in particular on Breiger et al. 2011, Breiger & Melamed 2013; Melamed et al. 2012, 2013). The POICN Database As Perliger and Pedahzur (2011) have noted, there has been a striking increase in efforts and resources invested in data collection on terrorist groups in recent years by academic and government agencies. Particularly notable in this respect have been the open-source, publicly available datasets maintained at the START Center at the University of Maryland, contributing to the present availability of high-resolution information (see also Hayden 2009). In this paper we make use of a new relational database developed by the START Center, focusing on terrorist plots, acquisitions, and attacks relating to CBRN agents from The Profiles of Incidents involving CBRN by Non-state actors (POICN) database distinguishes itself from other CBRN and terrorist attack databases by means of its transparent classification of source validity and inclusion of variables that rate the uncertainty that is sometimes present within and between sources. Explicitly incorporating and disclosing reliability and credibility levels allows for greater flexibility in tailoring the inclusion of cases for researchers specific 1

3 analytical requirements. In addition, the inclusion of such measures in this and similar datasets can facilitate more robust and defensible analyses and thus ultimately strengthen the role that social science can play in guiding and improving policy choices, especially in high-charged political and security contexts. When POICN is released publicly within the next several months, it will be the most comprehensive quantitative database on non-state actor CBRN events that have transpired within the past two decades. Transparent coding: source competence and objectivity; event credibility. In developing the POICN database, START researchers began with 499 potential cases from (a period subsequently extended to 2011), drawn from existing databases. The researchers rejected the widely applied but often-mistaken assumption that a case s inclusion in a dataset automatically equates to full validation of that case. A Source Evaluation Schema was developed, consisting of a set of operationalized variables and coding instructions. (See Ackerman and Pinson 2011 and Sawyer and Ackerman 2012 for more details.) Multiple variables focus on capturing the intentional and accidental distortion of information regarding CBRN activities. For each event, each source was coded for its competence. (For example, source competence was coded as questionable for those institutional publishers or authors generally known for high quality output, but where the particular source document describing an event exhibited prima facie indications of lower quality. Full competence was reserved for documents with no evident internal flaws produced by authors and institutions that have proven or researched competence in the geographical and substantive domain on which they are reporting, taking into account all that is known about their history and reputation as sources.) Likewise, each source was coded for its objectivity with respect to each event. (A newspaper that is generally measured in its approach to reporting but is known on occasion to take a very pro- Israeli or pro-palestinian stance on the Israeli-Palestinian issue is coded as evidencing only potential objectivity). The variable of credibility, coded once for each case (event) in the database, provides a measure of whether the event actually took place and whether that event really constituted a CBRN attack, based on corroboration between multiple independent sources. Sources coded as not competent or not objective were excluded from the credibility measure, which evaluates the number of remaining sources for each event with reference to their degree of mutual independence. A source is regarded as independent of another source if it does not share the same original authorship and does not rely on the same original source material. Source documents deriving from the same institution (such as the Associated Press) do not count as multiple sources. As indicated in Figure 1, the lowest level of credibility (35% of the events in the database) was assigned single-source events or events described by multiple non-independent sources. The intermediate level of credibility (14% of the events) was assigned to events reported by two independent sources not reflecting the same bias. The highest level of credibility (51% of 2

4 the events) was allocated to events for which there were three independent sources, or two independent sources with competing biases. INSERT FIGURE 1 ABOUT HERE Scope of the POICN database. We have highlighted the variables discussed above because they are directly relevant to the empirical analyses we report in this paper. Researchers will nonetheless want to know that the concern for validity and reliability of the events in the POICN database extends to many more design and coding features. For example, at least 54 of the key variables are coded for the presence / absence of both doubt and discrepancy, allowing researchers to assess in fine-grained ways effects of the level of validity of the data on the substantive conclusions drawn. The database contains records on 458 cases (events) spanning , and all of these 458 cases have been double-coded. POICN records 91 different groups and 75 lone actors as perpetrators in these events. Geospatial, longitudinal, technological, and organizational variables are coded and included. Basic descriptive analytics of POICN The POICN Dataset includes a spectrum of events ranging from protoplots (an event where a perpetrator is exploring CBRN pursuit but has not reached the level of plotting) to use of an agent (where a perpetrator employed or disseminated a CBRN substance in the commission of an attack ). Figure 2 shows the breakdown of these event types. Fifty-five percent of the 458 events are below the level of an actual attack. Due to the inherent uncertainty in reporting of CBRN events, two unique variables are coded: Event Uncertainty and Attack Uncertainty. Event Uncertainty is relevant when sources do not confirm whether or not a CBRN event actually occurred. Attack Uncertainty is relevant when sources do not confirm whether an event was actually an attack or was an accident or natural event. The great majority of events have no event or attack uncertainty (Figure 3). Events in POICN are coded for multiple agent types. Figure 4 shows that 78% of the events in POICN involve the pursuit or use of chemical weapons. Context and Research Questions INSERT FIGURES 2, 3, AND 4 ABOUT HERE Motivation. Given a data matrix (cases by variables), regression analysis as well as many of its generalizations may be thought of as the study of relations among the variables. With its typical assumption that the cases are a random sample representative of a population of interest, regression analysis makes the cases invisible, as Michael Shalev (2007) and other analysts of comparative politics have argued in their critiques of regression approaches. But often the cases are of interest, and the goal of the analysis should be to use the variables to let the cases be seen. Shalev (2007) discusses analyses where the cases are countries, and the 3

5 research agenda is comparative analysis of types of welfare states. In the example of the present paper, the cases are CBRN events, and our research agenda is comparative analysis of types of such events (discovering the types and how variables interact differently within each type). Moreover, in neither Shalev s examples nor those of the present paper could the analyst claim that the cases are a random sample. The POICN database aims at collecting all known cases of CBRN events within its date range, and there are surely dependencies among the events along multiple dimensions. (For example, two attacks attempted by the same group in adjacent months are likely not independent of each other. In addition, attacks using the same toxic agent by different groups but within the same country might well lack independence from one another. It seems quite limiting indeed to assume independence among all the cases.) We propose instead to discover regions of dependence among the cases on the basis of their attributes. Along with Shalev (2007), Charles Ragin (2008), and other researchers, we are willing to pay the costs of giving up our claim to significance testing in order to gain insight by more richly exploring the structuring internal to our dataset. Moreover, we show that we can do all this by deepening the framework within which regression analysis is conventionally understood. Outcome variable. We will assess the effects of several variables on the type of CBRN event. For this purpose we use only the 175 events taking place in the period that are coded at the highest credibility level (Figure 1). Previous research (Ackerman and Pinson, 2011) has demonstrated that the significance of variables in predictions of event type are affected by credibility level, and in this analysis we seek to generalize from only the most credible events. As mentioned above in discussion of Figure 2, the POICN database delineates eight different types of CBRN event, ranging from protoplots (e.g., knowledge that a terrorist group has hired a scientist with a CBRN specialty) all the way up to the use of an agent in an attack. We do not view these eight categories as forming an ordinal scale of intensity for CBRN activities. Rather, we separate the eight categories into two broad types: Type A = seeking a CBRN weapon = protoplots, plots, attempted acquisition, and possession of a non-weaponized agent. Type B = possessing a CBRN weapon = possession of a weapon, threat with possession, attempted use of weapon, and use of a CBRN substance in the commission of an attack. Table 1 provides more detail on each of the eight categories of our two broad types. The basic distinction is between seeking a CBRN weapon (Type A) and possessing one (Type B). TABLE 1 ABOUT HERE We will briefly elaborate our warning against any assumption that events of Type A in Table 1 (such as plots and acquisition attempts) are less serious than those of Type B (e.g., possession of a weaponized agent or its use in an actual attack). The weapon possessed in a Type 4

6 B event might be relatively crude (such as a small amount of radioactive material intended to be left in a building), and the attack might have caused no harm even though harm was intended. On the other hand, an event of Type A might be consequential if, for example, a Bulgarian businessman is approached by a contact with ties to al Qaeda and asked about the possibility of acquiring spent nuclear fuel rods (Event ID 23). To pursue this point: one of the variables in the POICN database is heightened interest, which is manifested by an event if any of three criteria are met (at least five casualties; involvement of a CBRN agent that is classified as a warfare agent; use of, or a plot to create, weaponization of the agent in at least a moderately sophisticated manner). When heightened interest is cross-classified with event type (for the 175 highcredibility events we use in our analysis, of which 6 events had missing data on heightened interest), we have: Event Type Heighted Interest Type A Type B No Yes Thus, whether an event involves possession or use of a weaponized agent (Type B) versus some form of plot, attempted acquisition, or possession of a non-weaponized agent (Type A) is only moderately, and non-significantly, related to heightened interest in the event (Yule s Q =.19, log(odds ratio) =.38; chi-square = 1.5 on 1 df). Both types of event (A and B) are potentially of great interest. As we seek to understand the qualities of these two broad event types (which partition the 175 events into two classes), we employ logistic regression. Predictor variables. We examine the effect on event type of nine predictors, all of which are (like the outcome variable) binary. Three of these pertain to world region: Russia and the NIS countries; the Middle East and North Africa; and South Asia. (The seven other regional categories used in the database serve as the omitted category. 1 ) Two pertain to weapon type: biological and chemical (with the omitted categories of radiological and nuclear providing a baseline). Four predictors pertain to the type of perpetrator: lone actors, religious extremists, cults, and ethnonationalist groups (with the three other classes of perpetrator providing the omitted-category baseline. 2 ) Descriptives are provided in Table 2. As seen there, 55% of the events pertain to the possession of a CBRN weapon (Type B events), while the remainder pertain to seeking such a weapon (Type A). Zero-order relationships and their calculation are illustrated at the bottom of 1 For 126 of the 175 events, a single country was listed. The remaining events were associated with multiple countries. For these, only the region of the first-listed country was coded. 2 Only 10 of the 175 events were coded as having multiple perpetrators, and only four events were coded as having multiple types of perpetrator. In these four cases only the first type was coded. 5

7 Table 2. Concerning biological weapons, for example, the odds are 12 / 25 = 0.48 that a biological weapon was possessed (Type B) versus sought (Type A). However, for non-biological weapons the odds are 84 / 54 = 1.56, implying that the odds on a Type B (rather than Type A) event are lower for biological weapons than for those of other types. Indeed, the odds ratio is (12 / 25) / (84 / 54) =.3086, and the log of the odds ratio is a negative (see Table 2). Bivariate relations of each predictor with the outcome are given similarly in Table 2. TABLE 2 ABOUT HERE Research questions. In a conventional multivariate study we would pose our central research question as follows: RQ1: What are the effects of the predictor variables on the outcome of possession and use of CBRN weapons (Type B events)? Indeed we are interested in that question. In addition, however, we are also interested in how we can use the variables to learn about the cases (the CBRN events). We thus also formulate the following non-traditional questions: RQ2: How can we compute the logistic regression coefficients as sums across the CBRN events? Therefore, how may we compute these same logistic regression coefficients as sums across clusters of events? RQ3: How can we use the clustering of cases (events) to discover interactions among the variables? And how can we use a single variable to induce clustering among the cases? RQ4: How can we define a network among the cases (CBRN events) such that that network yields the same predicted logits (log (p / (1 p)) as are produced by the standard logistic regression model? And how can this network among the cases provide us with a deeper understanding of the field of variables than is provided by the logistic regression equation? Logistic Regression Results The logistic regression model is reported in Table 3, and coefficients are given in both metric and standardized form. Events involving chemical agents, and event locations in Rusia NIS states, the Middle East and North Africa, and (especially) South Asia are all positively and significantly associated with events pertaining to CBRN weapon possession (Type B events). Religious extremist and ethnonationalist groups (the latter in a switch from the zero-order relationship reported in the previous table) are significantly and negatively associated with Type B events or, to say the same thing, significantly and positively associated with Type A events: the seeking of CBRN weapons. Events involving biological agents, and lone perpetrators and cult groups, are also negatively associated with Type B (positively associated with Type A), but not significantly so. 6

8 Turning Regression Modeling Inside Out TABLE 3 ABOUT HERE We will illustrate with the example of logistic regression, focusing on new insights that can be applied to the model application just reviewed. The logistic regression model implies log (p (1 p )) = X β where p is the modeled estimate of the probability that each event (in turn) is an event of Type B (possessing CBRN weapons), X is an events-by-variables design matrix for the predictor variables, and β is a vector of estimated logistic regression coefficients. We will compute the singular value decomposition (SVD) of matrix X: X = U S V T thus expressing the given events-by-predictor variables dataset (X) in terms of a set of orthogonal dimensions for the events (U) and a dual set orthogonal dimensions for the predictor variables (V), with S a diagonal matrix of weights (singular values) indicating the relative importance of each dimension. 3 (The superscript T signifies matrix transposition.) We will also define a diagonal matrix pertaining to the outcome variable, Y* = diag(log (p (1 p ))) Then, the usual formula for computing logistic regression coefficients is identical to β = V S 1 U T Y 1 where 1 is a vector of 1 s. The matrix product V S 1 U T Y is of dimension (predictor) variables by events, and the sum of its rows yields the identical logistic regression coefficients that are produced by standard packages. The novel element, however, is that the equation above indicates that the same logistic regression coefficients produced by the standard packages can be alternatively defined as sums across the cases, which in our example are the 175 events that we have been working with but that have remained invisible in Tables 2 and 3. In our previous work (e.g., Breiger et al. 2011, Melamed et al. 2013), we have examined regression coefficients as sums across clusters of the cases, where we have used inductive procedures (such as the k-means clustering algorithm) to identify the clusters. In this paper we take a different tack. We recognize that a binary variable may be simultaneously viewed as a clustering of cases into two clusters. Specifically, we distinguish less sophisticated events (such as a plot or an attack that involves leaving a radiological agent at a location, or contaminating 3 For convenience, we work not with X, the matrix of predictor variables, but with Z, in which each variable has been transformed to standard form (by subtracting its mean and dividing by its standard deviation). 7

9 drinking water with raw sewage) from highly sophisticated events (such as planning or constructing a containment system to protect perpetrators from the effects of the CBRN agent). 4 We partition our 175 CBRN events into two clusters: those that are low in sophistication, and those that are high. Using the equation above, we can write the same regression coefficients that we have previously computed (see the standardized logistic regression coefficients in the right-most column of Table 3) as sums across the 175 events partitioned into the lowsophistication and the high-sophistication clusters of events, and we have done precisely this in Table 4. 5 TABLE 4 AND FIGURE 5 ABOUT HERE Where do the numbers in Table 4 come from? This is illustrated in Figure 5 for the case of biological weapons. The logistic regression coefficient for bioweapons (in standardized form) is identical to the linear regression coefficient of the logit of p regressed on the residuals from bioweapons after regressing it on all the other predictor variables (see Figure 5). Moreover, the two clusters of cases have values on these variables (logit(p ) and residuals of the bioweapons variable) that sum to the logistic multiple regression coefficient,.0893 (given in standardized form). The astute reader might wonder whether the decomposition of the logistic regression coefficient (.0893) in Table 4, into a number pertaining to the cluster of low -sophistication cases (.1345) plus a number pertaining to the cluster of high -sophistication cases (.0451) implies that the latter numbers are the logistic regression coefficients estimated separately for the cases in each cluster. They are not, although the analogy is suggestive (as illustrated in Figure 5). We think of the latter two numbers instead as the versions of local logistic coefficients that are assumed by the logistic regression model that is applied across all the cases. Discovering statistical interactions. Notice the arrows at the right margin of Table 4, pointing to the variables chemical weapons and lone perpetrators. We observe that these variables move in opposite directions across the clusters. The chemical weapons variable decreases from.5656 (in the low sophistication cluster) to.2470 (in the high sophistication cluster). The lone perpetrator variable moves in the opposite direction, increasing from.2579 (in the low sophistication cluster) to.0576 (in the high sophistication cluster). We have shown that this opposite trending across clusters of cases can imply a statistical interaction among the implicated variables (Melamed et al. 2012, 2013). 6 Indeed, when we add an 4 To be precise, using the POICN variable for sophistication, we combine the categories of low and medium sophistication and contrast them with all other events. 5 There are two cases of rounding error in the fourth decimal place, but the two sets of regression coefficients (those in the right-most columns of Table 3 and Table 4) are otherwise identical. 6 The condition is that the cluster variable be significant when added to the (logistic) regression equation. In this case, the sophistication variable provides a statistically significant addition to the model of Table 3. 8

10 interaction term (chemical weapons lone perpetrator) to the logistic regression model of Table 3, we find a statistically significant improvement (Table 5). TABLE 5 ABOUT HERE The interpretation of the discovered statistical interaction is as follows: a lone perpetrator involved with chemicals is significantly more likely in an event that entails seeking weapons (Type A) rather than possessing or using them (Type B). The example provided here illustrates how clusters of cases (CBRN events) can be used to discover relations (statistical interactions) among variables. A network among events. Standard data-analysis packages produce estimates of log (p (1 p )) ; see the first equation on page 7 above. The identical logits may be computed in an alternative way by making use of the singular value decomposition given in the second equation on page 7. Making use of it, we may write, employing matrix multiplication: (U U T ) z* = log (p (1 p )) where U is from the above-mentioned singular value decomposition, and, with y containing the 0 s and 1 s of the observed dependent variable, z* = Xβ + (y - p ) The z* are analogous to the pseudo-values of logistic regression (compare Breiger et al. 2011, p. 29). The first equation in this section is important because it employs a network among the cases UU T to transform observed to fitted values of the outcome variable. 7 In order to produce the fitted logits, logistic regression (in effect) operates on a network of profile similarity among the cases. This network deserves serious study. The network we propose to study is cos(u), defined to have elements P ij / sqrt(p ii P jj ), where P = UU T. Thus, cos(u) is a network of cosines among all the rows of matrix U; it gives us cosines among all pairs of the 175 events. In order to gain visual clarity, we will report only the strong relations among the 175 events, defined as those with relatively high cosines above an arbitrarily chosen cutoff of Each of Figures 6, 7, and 8 shows (the identical) outcome network in its top panel, distinguishing between events of seeking to acquire CBRN materials (Type A) and possessing them (Type B events). Please recall that, by construction of the singular value decomposition (second equation on p. 7), no information on the outcome variable was taken into account in the formulation of network UU T. It is therefore remarkable that the (identical) top panel of each of 7 For this reason, this network appears in textbooks of mathematical statistics under the name, Projection Matrix. Data analysts concerned with regression diagnostics will know it as the Hat Matrix. 9

11 these figures distinguishes very sharply a substantial number of clusters that are fairly uniformly of one event type or the other. FIGURES 6, 7, 8 ABOUT HERE We have numbered the clusters in Figures 6 through 8 in the same order, and here we provide some overall comments about the organization of the network among events that (in a very direct sense) yields the estimated logits produced by the logistic regression model. Recall that the outcome variable, event type, distinguishes between seeking CBRN weapons (Type A) and possessing or using them (Type B). Other region refers to those outside of Russia and the NIS countries, South Asia, and the Middle East / North Africa. As the same event might entail multiple types of C, B, R, and/or N materials, we coded each event into a single type by emphasizing the less frequent category. 8 Inside this network, we may observe the following (compare Figs. 6-8): Cluster Number in Figs. 6-8 Characterization 1 Type A, , Russia, chemical 2 Type B, , other region, radiological & nuclear 3 Type A, [post-9/11], Middle East, chemical 5 other region, chemical 6 other region, biological 9 Type A, , South Asia, chemical 10 Mixed type, , other region, radiological 11 Type A, other region, chemical 12 Type B, , other region, biological As summarized just above, there is a great deal of patterning evident in Figs. 6-8; it is a patterning that organizes the production of estimated logits by the standard logistic regression model; however, it is a patterning that is invisible to conventional analyses using that model, because we are observing a network that exhibits no average relation between the variables. Cluster 10 is of particular interest because it provides a network bridge between two clusters of events in which chemicals are the weapon of choice, whereas Cluster 10 itself consists strongly of events organized around radiological weapons (Figure 8-b). The events of Cluster 10 exist in other regions (beyond those emphasized in the regression model of this paper; Fig. 7- b) and have largely taken place in the years (Fig. 6-b). 8 Thus we coded the event as biological if it was also chemical; as radiological if it was also biological or chemical; and as nuclear no matter what other materials might be involved. 10

12 Events located in Russia and the NIS countries (Cluster 1) are strongly associated with the earliest period covered in this database ( ), are strongly characterized by seeking rather than possessing weapons, and are centered around chemical weapons. Identification of Cluster 2 seems a worthwhile result of our network procedures. Events in Cluster 2 are uniquely centered on radiological and nuclear materials, focused on possession (Type B) rather than seeking, and concentrated in the period These events are strongly located in regions other than the three emphasized in our logistic regression model. Events centered in the Middle East and North Africa (Cluster 3) are strongly associated with the entire post-9/11 period, are focused on seeking (Type A events), and center on chemical weapons. Events in South Asia (Cluster 9) are highly interrelated ( dense ), concentrated in the most recent period (post-2006), focused on seeking and on chemical weapons. Clusters 6 and 12 are strikingly strong in focusing on biological events. Both clusters concentrate in regions other than those emphasized in the regression modeling, but Cluster 6 was of mixed event type whereas Cluster 12 events tell a story of possession and/or use (Type B). Discussion This paper has pursued at length a line of criticism of logistic regression modeling that is also taken up (albeit in a much different though fascinating way) in the paper of Bear Braumoeller written for the same panel. Specifically, a purely additive specification in a logit (or related) regression model will not capture the intuition that a causal effect will vary depending on the values of other variables (Braumoeller 2013, p. 5). Moreover, as Braumoeller also points out, interaction terms in logistic regression models are problematic (in part) because the number of interactions required typically renders interpretation difficult or (often) impossible (Braumoeller 2013, p. 5). And, in a very particular and partial way, this paper attempts to address this problem by having the courage to move quantitative approaches more toward description of what Philip Schrodt, another panel speaker, refers to as the new events datasets that seem to require a more qualitative orientation emphasizing thick description (Schrodt 2012, p. 551) as well as more attention from the point of view of network analysis (Schrodt 2012, p. 556). To cite one example: a large number of (potential) statistical interactions are examined simultaneously by the decomposition of regression coefficients in Table 4 of this paper. The basic premise of the modeling effort in this paper is that although regression modeling begins with a data matrix of dimension cases variables, all standard modeling proceeds quickly to ignore the cases in order to focus on the variables. Moreover, there is a useful dual to the standard regression approach, one in which the usual regression coefficients are computed as sums across cases (and, hence, clusters of cases), an alternative in which new binary variables can be analyzed directly as the clusters that they are and used to discover 11

13 statistical interactions among the variables, and an approach in which a network of profile similarity among the cases is seen to be implicated every time an analyst hits the compute regression button on any favorite statistical package. In pursing these modeling ideas, we had the superbly good fortune to be working with the most serious open-source database on CBRN events that has ever been constructed. Among the newly-opened possibilities, we want to reconsider the results reported here confined to the cases of highest credibility with respect to their robustness across a larger number of at least moderately credible events. In ways that this paper has just begun to tap, the POICN database will have a huge impact on future scientific analysis of CBRN as an arena for conflict. References Ackerman, Gary A., and Lauren Pinson Speaking Truth to Sources: Introducing a Method for the Quantitative Evaluation of Open-Sources in Event Data. College Park, MD: National Consortium for the Study of Terrorism and Responses to Terrorism (START Center): working paper. Asal, V.H., and R.K. Rethemeyer "Islamist use and Pursuit of CBRN Terrorism." Pp in Jihadists and Weapons of Mass Destruction, edited by G. Ackerman, and J. Tamsett. Boca Raton, FL and London: CRC Press. Asal, Victor, and R. Karl Rethemeyer "The Nature of the Beast: Organizational Structures and the Lethality of Terrorist Attacks." Journal of Politics 70 (2): Braumoeller, Bear F The Anna Karenina Principle in International Relations. Paper prepared for presentation at the Annual Meeting of the International Studies Association, San Francisco (April). Breiger, R.L., G.A. Ackerman, V. Asal, D. Melamed, H.B. Milward, R.K. Rethemeyer, and E. Schoon "Application of a Profile Similarity Methodology for Identifying Terrorist Groups that use Or Pursue CBRN Weapons." Pp in Social Computing, Behavioral- Cultural Modeling and Prediction, edited by J. Salerno, S.J. Yang, D. Nau, and S. Chai. Berlin;Heidelberg: Springer. Breiger, R.L., and David Melamed The Duality of Organizations and their Attributes: Turning Regression Modeling Inside Out. Research in the Sociology of Organizations: in press. Hayden, N.K "Terrifying Landscapes: Understanding Motivations of Non-State Actors to Acquire and/or use Weapons of Mass Destruction." Pp in Unconventional 12

14 Weapons and International Terrorism: Challenges and New Approaches, edited by M. Ranstorp, and M. Normark. London and New York: Routledge. Melamed, David, Ronald L. Breiger, and Eric Schoon "The Duality of Clusters and Statistical Interactions." Sociological Methods & Research 42 (1): Melamed, D., E. Schoon, R. Breiger, V. Asal, and R.K. Rethemeyer "Using Organizational Similarity to Identify Statistical Interactions for Improving Situational Awareness of CBRN Activities." Pp in Social Computing, Behavioral-Cultural Modeling, and Prediction (Lecture Notes in Computer Science 7227), edited by S.J. Yang, A.M. Greenberg, and M. Endsley. Berlin; Heidelberg: Springer-Verlag. Perliger, Arie, and Ami Pedahzur "Social Network Analysis in the Study of Terrorism and Political Violence." PS: Political Science & Politics 44 (01): Ragin, Charles C Redesigning Social Inquiry: Fuzzy Sets and Beyond. Chicago and London: University of Chicago Press. Sawyer, John, and Gary Ackerman Promethean Journeys: Examining the Mechanisms by Which Terrorists Acquire New Technologies of Lethality. Paper presented at the Annual Meeting of the International Studies Association, San Diego. Schrodt, Philip A "Precedents, Progress, and Prospects in Political Event Data." International Interactions 38 (4): Shalev, M "Limits and Alternatives to Multiple Regression in Comparative Research." Pp in Comparative Social Research (Symposium on Methodology in Comparative Research), edited by L. Mjøset, and T.H. Clausen. Elsevier. 13

15 Figure 1. Levels of Event Credibility Level 1: single source or multiple non-independent sources Level 2: two independent sources Level 3: 3+ independent sources, or 2 with competing bias 14

16 CBRN Terrorist Events by Event Type, % 8% 15% 11% 3% 6% 3% 45% Use of Agent Attempted Use Threat with Possession Acquisition of a Weapon Acquisition of an Agent Attempted Acquisition Plot Protoplot Figure 2. CBRN Terrorist Events by Event Type,

17 a) Event Uncertainty 6% 16% 78% No Uncertainty Some Uncertainty High Uncertainty b) Attack Uncertainty 11% 4% 85% Figure 3. Event Uncertainty (a) and Attack Uncertainty (b) 16

18 CBRN Terrorist Events by Agent Type, % % 10.8% 3.7% Chemical Biological Radiological Nuclear Figure 4. CBRN Events by Agent Type,

19 Figure 5. The X-axis is residuals of biological weapon variable regressed linearly against all other predictor variables. The Y-axis is log (p (1 p )). The linear regression line shown in the Figure has slope.0893, identical to the (standardized) logistic regression coefficient for bioweapons in Table 3. The 175 events are depicted as open red squares ( low sophistication ) or as closed blue circles ( high sophistication ). We may compute X s Y s s X t 2 t where subscript t indexes all 175 cases, and subscript s indexes cases in a subset. When s indexes the low-sophistication cases, the above expression equals When s indexes the high-sophistication cases, the expression equals When s indexes all 175 cases (when s = t), the expression equals These are the three numbers given for bioweapons in Table 4. 18

20 a) cos(u) >.60; events of Type A (red circles) and Type B (blue squares) b) cos(u) >.60, events by Year: (yellow), (light green), (dark green); circles for Type A, squares for Type B Figure 6. cos(u) >.60, showing Event Type (a) and Year of occurrence (b) 19

21 a) cos(u) >.60; events of Type A (red circles) and Type B (blue squares) b) cos(u) >.60; events in Russia (blue), South Asia (green), Middle East (red) Figure 7. cos(u) >.60, showing Event Type (a) and Region of occurrence (b) 20

22 a) cos(u) >.60; events of Type A (red circles) and Type B (blue squares) b) cos(u) >.60; CBRN type Chem (tan), Bio (green), Rad (blue), Nuc (red) Figure 8. cos(u) >.60, showing Event Type (a) and CBRN material (b) 21

23 Table 1. The Two Broad Types (Seeking and Possessing) of CBRN Event* Event Type A: Seeking a CBRN Weapon Protoplot Plot The sources do not present evidence of an actual plot, bur rather mention events that may lay the groundwork for an actual plot. Example: Knowledge of a terrorist group hiring a scientist with a CBRN weapons specialty. The perpetrator(s) seriously considered acquiring and using CBRN materials as a weapon, but did not make an attempt to acquire the agent. Attempted acquisition Possession of a nonweaponized agent There is evidence to suggest that the perpetrator(s) attempted to acquire a CBRN substance for use as a weapon, but no evidence of success. Includes the attempted (but unsuccessful or abandoned) acquisition of raw materials or an intact CBRN weapon. If the sole terrorist organization involved in an event is the intended recipient of an agent /weapon that was intercepted en route, the event may be coded as an attempted acquisition with the terrorist organization as the perpetrator. The perpetrator(s) suceed in possessing a CBRN agent but this agent does not constitute a weapon (e.g., it is not in a deliverable form, lacking an effective delivery mechanism for the intended attack). Event Type B: Possession or Use of a Weaponized Agent Possession of a weapon Threat with possession Attempted use of agent Use of agent The perpetrator(s) posessed both the agent and delivery mechanism in a form that constitutes a viable weapon or can easily be assembled into such a weapon at the time of reporting. The completed weapon may be crude (e.g., radioactive material the perpetrator plans to leave in a building) if there was evidence that the perpetrator intended to use the weapon in this crude form. The perpetrator(s) both threatened to use a CBRN substance and actually had the weapon in their possession at the time fo the threat. The perpetrator(s) attempted to employ or disseminate a CBRN substance but no agent was actually released. The perpetrator(s) employed or disseminated a CBRN substance in the commission of an attack. If a small amount of agent was used, even if no harm was caused, it is coded as "use of agent" unless there is proof the event was not meant to cause harm. *Note: Type A comprises 45.1%, and Type B 54.9%, of the 175 events we analyze. 22

24 Table 2. One-variable marginals, 2-variable relations with the Outcome variable Possession of a CBRN weapon (Type B) Fraction of events Log odds Overall 96 / 175 = 55% Log odds ratio Independent variables Russia, NIS 12/15 = 80% Middle East 16/27 = 59% South Asia 16/18 = 89% biological weapon 12/37 = 32% chemical weapon 83/123 = 67% lone actor 19/38 = 50% religious extremists 33/72 = 46% cults 1/4 = 25% ethnonationalist 16/25 = 64% Example of calculations: For "biological weapon," we have: Bio-weapon? Possession? No Yes No Yes / ( ) = 32% log(12/25) =.734 log((12/25) / (84/54)) =

25 Table 3. The Logistic Regression Model (Outcome is whether each event is of Type B) Both metric coefficients and standardized coefficients are reported. Unstandardized Coefficients Standardized Coefs. Estimate Std. Error t value Pr(> t ) Estimate (Intercept) Russia, NIS * * Middle East * * South Asia ** * biological weapon chemical weapon ** * lone actor religious extremists ** * cults ethnonationalist * * Assessment of model fit: value df Null deviance: Residual deviance: Reduction: ** * ** * ** * 24

26 Table 4. The usual regression coefficients (see the right-most column in Table 3) are sums across clusters of events! Sophistication: Variable: Low High Sum (Intercept) Russia, NIS Middle East South Asia biological weapon chemical weapon lone actor religious extremists cults ethnonationalist

27 Table 5. The model of Table 3 with the addition of a (chemical weapons * lone actor) interaction Unstandardized Coefficients Standardized Coefs. Estimate Std. Error t value Pr(> t ) Estimate (Intercept) Russia, NIS * * Middle East South Asia ** ** biological weapon chemical weapon *** *** lone actor religious extremists *** *** cults * * ethnonationalist ** ** chem * lone actor ** ** Assessment of model fit: value df Null deviance: Residual deviance: Reduction:

MULTIPLE LINEAR REGRESSION 24.1 INTRODUCTION AND OBJECTIVES OBJECTIVES

MULTIPLE LINEAR REGRESSION 24.1 INTRODUCTION AND OBJECTIVES OBJECTIVES 24 MULTIPLE LINEAR REGRESSION 24.1 INTRODUCTION AND OBJECTIVES In the previous chapter, simple linear regression was used when you have one independent variable and one dependent variable. This chapter

More information

MEASURES OF ASSOCIATION AND REGRESSION

MEASURES OF ASSOCIATION AND REGRESSION DEPARTMENT OF POLITICAL SCIENCE AND INTERNATIONAL RELATIONS Posc/Uapp 816 MEASURES OF ASSOCIATION AND REGRESSION I. AGENDA: A. Measures of association B. Two variable regression C. Reading: 1. Start Agresti

More information

ATTACK MODALITIES. Draft 1 Charles P. Blair

ATTACK MODALITIES. Draft 1 Charles P. Blair ATTACK MODALITIES Draft 1 Charles P. Blair ATTACK MODALITIES Attack Modalities refers to the actual methods and techniques that terrorists choose to employ to attack particular targets. There are several

More information

You must answer question 1.

You must answer question 1. Research Methods and Statistics Specialty Area Exam October 28, 2015 Part I: Statistics Committee: Richard Williams (Chair), Elizabeth McClintock, Sarah Mustillo You must answer question 1. 1. Suppose

More information

Statistical reports Regression, 2010

Statistical reports Regression, 2010 Statistical reports Regression, 2010 Niels Richard Hansen June 10, 2010 This document gives some guidelines on how to write a report on a statistical analysis. The document is organized into sections that

More information

Research Prospectus. Your major writing assignment for the quarter is to prepare a twelve-page research prospectus.

Research Prospectus. Your major writing assignment for the quarter is to prepare a twelve-page research prospectus. Department of Political Science UNIVERSITY OF CALIFORNIA, SAN DIEGO Philip G. Roeder Research Prospectus Your major writing assignment for the quarter is to prepare a twelve-page research prospectus. A

More information

Analysis of Environmental Data Conceptual Foundations: En viro n m e n tal Data

Analysis of Environmental Data Conceptual Foundations: En viro n m e n tal Data Analysis of Environmental Data Conceptual Foundations: En viro n m e n tal Data 1. Purpose of data collection...................................................... 2 2. Samples and populations.......................................................

More information

11/18/2013. Correlational Research. Correlational Designs. Why Use a Correlational Design? CORRELATIONAL RESEARCH STUDIES

11/18/2013. Correlational Research. Correlational Designs. Why Use a Correlational Design? CORRELATIONAL RESEARCH STUDIES Correlational Research Correlational Designs Correlational research is used to describe the relationship between two or more naturally occurring variables. Is age related to political conservativism? Are

More information

Speaker Notes: Qualitative Comparative Analysis (QCA) in Implementation Studies

Speaker Notes: Qualitative Comparative Analysis (QCA) in Implementation Studies Speaker Notes: Qualitative Comparative Analysis (QCA) in Implementation Studies PART 1: OVERVIEW Slide 1: Overview Welcome to Qualitative Comparative Analysis in Implementation Studies. This narrated powerpoint

More information

CHAPTER 3 RESEARCH METHODOLOGY

CHAPTER 3 RESEARCH METHODOLOGY CHAPTER 3 RESEARCH METHODOLOGY 3.1 Introduction 3.1 Methodology 3.1.1 Research Design 3.1. Research Framework Design 3.1.3 Research Instrument 3.1.4 Validity of Questionnaire 3.1.5 Statistical Measurement

More information

3 CONCEPTUAL FOUNDATIONS OF STATISTICS

3 CONCEPTUAL FOUNDATIONS OF STATISTICS 3 CONCEPTUAL FOUNDATIONS OF STATISTICS In this chapter, we examine the conceptual foundations of statistics. The goal is to give you an appreciation and conceptual understanding of some basic statistical

More information

bivariate analysis: The statistical analysis of the relationship between two variables.

bivariate analysis: The statistical analysis of the relationship between two variables. bivariate analysis: The statistical analysis of the relationship between two variables. cell frequency: The number of cases in a cell of a cross-tabulation (contingency table). chi-square (χ 2 ) test for

More information

Analysis of the Reliability and Validity of an Edgenuity Algebra I Quiz

Analysis of the Reliability and Validity of an Edgenuity Algebra I Quiz Analysis of the Reliability and Validity of an Edgenuity Algebra I Quiz This study presents the steps Edgenuity uses to evaluate the reliability and validity of its quizzes, topic tests, and cumulative

More information

Business Statistics Probability

Business Statistics Probability Business Statistics The following was provided by Dr. Suzanne Delaney, and is a comprehensive review of Business Statistics. The workshop instructor will provide relevant examples during the Skills Assessment

More information

6. Unusual and Influential Data

6. Unusual and Influential Data Sociology 740 John ox Lecture Notes 6. Unusual and Influential Data Copyright 2014 by John ox Unusual and Influential Data 1 1. Introduction I Linear statistical models make strong assumptions about the

More information

Describe what is meant by a placebo Contrast the double-blind procedure with the single-blind procedure Review the structure for organizing a memo

Describe what is meant by a placebo Contrast the double-blind procedure with the single-blind procedure Review the structure for organizing a memo Business Statistics The following was provided by Dr. Suzanne Delaney, and is a comprehensive review of Business Statistics. The workshop instructor will provide relevant examples during the Skills Assessment

More information

CHAPTER ONE CORRELATION

CHAPTER ONE CORRELATION CHAPTER ONE CORRELATION 1.0 Introduction The first chapter focuses on the nature of statistical data of correlation. The aim of the series of exercises is to ensure the students are able to use SPSS to

More information

Dealing with Missing Data in Group-Level Studies of Terrorism

Dealing with Missing Data in Group-Level Studies of Terrorism Dealing with Missing Data in Group-Level Studies of Terrorism Bryan Arva and John Beieler Pennsylvania State University September 30, 2014 Abstract One prominent area of research in terrorism studies focuses

More information

How to Create Better Performing Bayesian Networks: A Heuristic Approach for Variable Selection

How to Create Better Performing Bayesian Networks: A Heuristic Approach for Variable Selection How to Create Better Performing Bayesian Networks: A Heuristic Approach for Variable Selection Esma Nur Cinicioglu * and Gülseren Büyükuğur Istanbul University, School of Business, Quantitative Methods

More information

Causal Mediation Analysis with the CAUSALMED Procedure

Causal Mediation Analysis with the CAUSALMED Procedure Paper SAS1991-2018 Causal Mediation Analysis with the CAUSALMED Procedure Yiu-Fai Yung, Michael Lamm, and Wei Zhang, SAS Institute Inc. Abstract Important policy and health care decisions often depend

More information

MMI 409 Spring 2009 Final Examination Gordon Bleil. 1. Is there a difference in depression as a function of group and drug?

MMI 409 Spring 2009 Final Examination Gordon Bleil. 1. Is there a difference in depression as a function of group and drug? MMI 409 Spring 2009 Final Examination Gordon Bleil Table of Contents Research Scenario and General Assumptions Questions for Dataset (Questions are hyperlinked to detailed answers) 1. Is there a difference

More information

CRITERIA FOR USE. A GRAPHICAL EXPLANATION OF BI-VARIATE (2 VARIABLE) REGRESSION ANALYSISSys

CRITERIA FOR USE. A GRAPHICAL EXPLANATION OF BI-VARIATE (2 VARIABLE) REGRESSION ANALYSISSys Multiple Regression Analysis 1 CRITERIA FOR USE Multiple regression analysis is used to test the effects of n independent (predictor) variables on a single dependent (criterion) variable. Regression tests

More information

SUMMER 2011 RE-EXAM PSYF11STAT - STATISTIK

SUMMER 2011 RE-EXAM PSYF11STAT - STATISTIK SUMMER 011 RE-EXAM PSYF11STAT - STATISTIK Full Name: Årskortnummer: Date: This exam is made up of three parts: Part 1 includes 30 multiple choice questions; Part includes 10 matching questions; and Part

More information

Section 6: Analysing Relationships Between Variables

Section 6: Analysing Relationships Between Variables 6. 1 Analysing Relationships Between Variables Section 6: Analysing Relationships Between Variables Choosing a Technique The Crosstabs Procedure The Chi Square Test The Means Procedure The Correlations

More information

Section 3.2 Least-Squares Regression

Section 3.2 Least-Squares Regression Section 3.2 Least-Squares Regression Linear relationships between two quantitative variables are pretty common and easy to understand. Correlation measures the direction and strength of these relationships.

More information

Still important ideas

Still important ideas Readings: OpenStax - Chapters 1 13 & Appendix D & E (online) Plous Chapters 17 & 18 - Chapter 17: Social Influences - Chapter 18: Group Judgments and Decisions Still important ideas Contrast the measurement

More information

Regression Discontinuity Analysis

Regression Discontinuity Analysis Regression Discontinuity Analysis A researcher wants to determine whether tutoring underachieving middle school students improves their math grades. Another wonders whether providing financial aid to low-income

More information

One-Way Independent ANOVA

One-Way Independent ANOVA One-Way Independent ANOVA Analysis of Variance (ANOVA) is a common and robust statistical test that you can use to compare the mean scores collected from different conditions or groups in an experiment.

More information

DAZED AND CONFUSED: THE CHARACTERISTICS AND BEHAVIOROF TITLE CONFUSED READERS

DAZED AND CONFUSED: THE CHARACTERISTICS AND BEHAVIOROF TITLE CONFUSED READERS Worldwide Readership Research Symposium 2005 Session 5.6 DAZED AND CONFUSED: THE CHARACTERISTICS AND BEHAVIOROF TITLE CONFUSED READERS Martin Frankel, Risa Becker, Julian Baim and Michal Galin, Mediamark

More information

Unit 1 Exploring and Understanding Data

Unit 1 Exploring and Understanding Data Unit 1 Exploring and Understanding Data Area Principle Bar Chart Boxplot Conditional Distribution Dotplot Empirical Rule Five Number Summary Frequency Distribution Frequency Polygon Histogram Interquartile

More information

Preliminary Report on Simple Statistical Tests (t-tests and bivariate correlations)

Preliminary Report on Simple Statistical Tests (t-tests and bivariate correlations) Preliminary Report on Simple Statistical Tests (t-tests and bivariate correlations) After receiving my comments on the preliminary reports of your datasets, the next step for the groups is to complete

More information

Performance of Median and Least Squares Regression for Slightly Skewed Data

Performance of Median and Least Squares Regression for Slightly Skewed Data World Academy of Science, Engineering and Technology 9 Performance of Median and Least Squares Regression for Slightly Skewed Data Carolina Bancayrin - Baguio Abstract This paper presents the concept of

More information

CSE 255 Assignment 9

CSE 255 Assignment 9 CSE 255 Assignment 9 Alexander Asplund, William Fedus September 25, 2015 1 Introduction In this paper we train a logistic regression function for two forms of link prediction among a set of 244 suspected

More information

MEA DISCUSSION PAPERS

MEA DISCUSSION PAPERS Inference Problems under a Special Form of Heteroskedasticity Helmut Farbmacher, Heinrich Kögel 03-2015 MEA DISCUSSION PAPERS mea Amalienstr. 33_D-80799 Munich_Phone+49 89 38602-355_Fax +49 89 38602-390_www.mea.mpisoc.mpg.de

More information

Modelling Research Productivity Using a Generalization of the Ordered Logistic Regression Model

Modelling Research Productivity Using a Generalization of the Ordered Logistic Regression Model Modelling Research Productivity Using a Generalization of the Ordered Logistic Regression Model Delia North Temesgen Zewotir Michael Murray Abstract In South Africa, the Department of Education allocates

More information

Studying the effect of change on change : a different viewpoint

Studying the effect of change on change : a different viewpoint Studying the effect of change on change : a different viewpoint Eyal Shahar Professor, Division of Epidemiology and Biostatistics, Mel and Enid Zuckerman College of Public Health, University of Arizona

More information

WELCOME! Lecture 11 Thommy Perlinger

WELCOME! Lecture 11 Thommy Perlinger Quantitative Methods II WELCOME! Lecture 11 Thommy Perlinger Regression based on violated assumptions If any of the assumptions are violated, potential inaccuracies may be present in the estimated regression

More information

Modeling Terrorist Beliefs and Motivations

Modeling Terrorist Beliefs and Motivations CREATE Research Archive Research Project Summaries 2010 Modeling Terrorist Beliefs and Motivations Richard John University of Southern California, richardj@usc.edu Heather Rosoff University of Southern

More information

(CORRELATIONAL DESIGN AND COMPARATIVE DESIGN)

(CORRELATIONAL DESIGN AND COMPARATIVE DESIGN) UNIT 4 OTHER DESIGNS (CORRELATIONAL DESIGN AND COMPARATIVE DESIGN) Quasi Experimental Design Structure 4.0 Introduction 4.1 Objectives 4.2 Definition of Correlational Research Design 4.3 Types of Correlational

More information

Addendum: Multiple Regression Analysis (DRAFT 8/2/07)

Addendum: Multiple Regression Analysis (DRAFT 8/2/07) Addendum: Multiple Regression Analysis (DRAFT 8/2/07) When conducting a rapid ethnographic assessment, program staff may: Want to assess the relative degree to which a number of possible predictive variables

More information

Political Science 15, Winter 2014 Final Review

Political Science 15, Winter 2014 Final Review Political Science 15, Winter 2014 Final Review The major topics covered in class are listed below. You should also take a look at the readings listed on the class website. Studying Politics Scientifically

More information

Still important ideas

Still important ideas Readings: OpenStax - Chapters 1 11 + 13 & Appendix D & E (online) Plous - Chapters 2, 3, and 4 Chapter 2: Cognitive Dissonance, Chapter 3: Memory and Hindsight Bias, Chapter 4: Context Dependence Still

More information

CHAPTER VI RESEARCH METHODOLOGY

CHAPTER VI RESEARCH METHODOLOGY CHAPTER VI RESEARCH METHODOLOGY 6.1 Research Design Research is an organized, systematic, data based, critical, objective, scientific inquiry or investigation into a specific problem, undertaken with the

More information

How to interpret results of metaanalysis

How to interpret results of metaanalysis How to interpret results of metaanalysis Tony Hak, Henk van Rhee, & Robert Suurmond Version 1.0, March 2016 Version 1.3, Updated June 2018 Meta-analysis is a systematic method for synthesizing quantitative

More information

11/24/2017. Do not imply a cause-and-effect relationship

11/24/2017. Do not imply a cause-and-effect relationship Correlational research is used to describe the relationship between two or more naturally occurring variables. Is age related to political conservativism? Are highly extraverted people less afraid of rejection

More information

Sawtooth Software. The Number of Levels Effect in Conjoint: Where Does It Come From and Can It Be Eliminated? RESEARCH PAPER SERIES

Sawtooth Software. The Number of Levels Effect in Conjoint: Where Does It Come From and Can It Be Eliminated? RESEARCH PAPER SERIES Sawtooth Software RESEARCH PAPER SERIES The Number of Levels Effect in Conjoint: Where Does It Come From and Can It Be Eliminated? Dick Wittink, Yale University Joel Huber, Duke University Peter Zandan,

More information

THE STATSWHISPERER. Introduction to this Issue. Doing Your Data Analysis INSIDE THIS ISSUE

THE STATSWHISPERER. Introduction to this Issue. Doing Your Data Analysis INSIDE THIS ISSUE Spring 20 11, Volume 1, Issue 1 THE STATSWHISPERER The StatsWhisperer Newsletter is published by staff at StatsWhisperer. Visit us at: www.statswhisperer.com Introduction to this Issue The current issue

More information

Data and Statistics 101: Key Concepts in the Collection, Analysis, and Application of Child Welfare Data

Data and Statistics 101: Key Concepts in the Collection, Analysis, and Application of Child Welfare Data TECHNICAL REPORT Data and Statistics 101: Key Concepts in the Collection, Analysis, and Application of Child Welfare Data CONTENTS Executive Summary...1 Introduction...2 Overview of Data Analysis Concepts...2

More information

Chapter 02. Basic Research Methodology

Chapter 02. Basic Research Methodology Chapter 02 Basic Research Methodology Definition RESEARCH Research is a quest for knowledge through diligent search or investigation or experimentation aimed at the discovery and interpretation of new

More information

PEER REVIEW HISTORY ARTICLE DETAILS VERSION 1 - REVIEW. Ball State University

PEER REVIEW HISTORY ARTICLE DETAILS VERSION 1 - REVIEW. Ball State University PEER REVIEW HISTORY BMJ Open publishes all reviews undertaken for accepted manuscripts. Reviewers are asked to complete a checklist review form (see an example) and are provided with free text boxes to

More information

Examining differences between two sets of scores

Examining differences between two sets of scores 6 Examining differences between two sets of scores In this chapter you will learn about tests which tell us if there is a statistically significant difference between two sets of scores. In so doing you

More information

Preliminary Conclusion

Preliminary Conclusion 1 Exploring the Genetic Component of Political Participation Brad Verhulst Virginia Institute for Psychiatric and Behavioral Genetics Virginia Commonwealth University Theories of political participation,

More information

Describe what is meant by a placebo Contrast the double-blind procedure with the single-blind procedure Review the structure for organizing a memo

Describe what is meant by a placebo Contrast the double-blind procedure with the single-blind procedure Review the structure for organizing a memo Please note the page numbers listed for the Lind book may vary by a page or two depending on which version of the textbook you have. Readings: Lind 1 11 (with emphasis on chapters 10, 11) Please note chapter

More information

Answers to end of chapter questions

Answers to end of chapter questions Answers to end of chapter questions Chapter 1 What are the three most important characteristics of QCA as a method of data analysis? QCA is (1) systematic, (2) flexible, and (3) it reduces data. What are

More information

CHAPTER TWO REGRESSION

CHAPTER TWO REGRESSION CHAPTER TWO REGRESSION 2.0 Introduction The second chapter, Regression analysis is an extension of correlation. The aim of the discussion of exercises is to enhance students capability to assess the effect

More information

Estimating Heterogeneous Choice Models with Stata

Estimating Heterogeneous Choice Models with Stata Estimating Heterogeneous Choice Models with Stata Richard Williams Notre Dame Sociology rwilliam@nd.edu West Coast Stata Users Group Meetings October 25, 2007 Overview When a binary or ordinal regression

More information

A review of statistical methods in the analysis of data arising from observer reliability studies (Part 11) *

A review of statistical methods in the analysis of data arising from observer reliability studies (Part 11) * A review of statistical methods in the analysis of data arising from observer reliability studies (Part 11) * by J. RICHARD LANDIS** and GARY G. KOCH** 4 Methods proposed for nominal and ordinal data Many

More information

BIOL 458 BIOMETRY Lab 7 Multi-Factor ANOVA

BIOL 458 BIOMETRY Lab 7 Multi-Factor ANOVA BIOL 458 BIOMETRY Lab 7 Multi-Factor ANOVA PART 1: Introduction to Factorial ANOVA ingle factor or One - Way Analysis of Variance can be used to test the null hypothesis that k or more treatment or group

More information

Social Effects in Blau Space:

Social Effects in Blau Space: Social Effects in Blau Space: Miller McPherson and Jeffrey A. Smith Duke University Abstract We develop a method of imputing characteristics of the network alters of respondents in probability samples

More information

THE USE OF MULTIVARIATE ANALYSIS IN DEVELOPMENT THEORY: A CRITIQUE OF THE APPROACH ADOPTED BY ADELMAN AND MORRIS A. C. RAYNER

THE USE OF MULTIVARIATE ANALYSIS IN DEVELOPMENT THEORY: A CRITIQUE OF THE APPROACH ADOPTED BY ADELMAN AND MORRIS A. C. RAYNER THE USE OF MULTIVARIATE ANALYSIS IN DEVELOPMENT THEORY: A CRITIQUE OF THE APPROACH ADOPTED BY ADELMAN AND MORRIS A. C. RAYNER Introduction, 639. Factor analysis, 639. Discriminant analysis, 644. INTRODUCTION

More information

The Regression-Discontinuity Design

The Regression-Discontinuity Design Page 1 of 10 Home» Design» Quasi-Experimental Design» The Regression-Discontinuity Design The regression-discontinuity design. What a terrible name! In everyday language both parts of the term have connotations

More information

Exemplar for Internal Assessment Resource Mathematics Level 3. Resource title: Sport Science. Investigate bivariate measurement data

Exemplar for Internal Assessment Resource Mathematics Level 3. Resource title: Sport Science. Investigate bivariate measurement data Exemplar for internal assessment resource Mathematics 3.9A for Achievement Standard 91581 Exemplar for Internal Assessment Resource Mathematics Level 3 Resource title: Sport Science This exemplar supports

More information

Agents with Attitude: Exploring Coombs Unfolding Technique with Agent-Based Models

Agents with Attitude: Exploring Coombs Unfolding Technique with Agent-Based Models Int J Comput Math Learning (2009) 14:51 60 DOI 10.1007/s10758-008-9142-6 COMPUTER MATH SNAPHSHOTS - COLUMN EDITOR: URI WILENSKY* Agents with Attitude: Exploring Coombs Unfolding Technique with Agent-Based

More information

Chapter Eight: Multivariate Analysis

Chapter Eight: Multivariate Analysis Chapter Eight: Multivariate Analysis Up until now, we have covered univariate ( one variable ) analysis and bivariate ( two variables ) analysis. We can also measure the simultaneous effects of two or

More information

Assignment Research Article (in Progress)

Assignment Research Article (in Progress) Political Science 225 UNIVERSITY OF CALIFORNIA, SAN DIEGO Assignment Research Article (in Progress) Philip G. Roeder Your major writing assignment for the quarter is to prepare a twelve-page research article

More information

THIS PROBLEM HAS BEEN SOLVED BY USING THE CALCULATOR. A 90% CONFIDENCE INTERVAL IS ALSO SHOWN. ALL QUESTIONS ARE LISTED BELOW THE RESULTS.

THIS PROBLEM HAS BEEN SOLVED BY USING THE CALCULATOR. A 90% CONFIDENCE INTERVAL IS ALSO SHOWN. ALL QUESTIONS ARE LISTED BELOW THE RESULTS. Math 117 Confidence Intervals and Hypothesis Testing Interpreting Results SOLUTIONS The results are given. Interpret the results and write the conclusion within context. Clearly indicate what leads to

More information

VULNERABILITY AND EXPOSURE TO CRIME: APPLYING RISK TERRAIN MODELING

VULNERABILITY AND EXPOSURE TO CRIME: APPLYING RISK TERRAIN MODELING VULNERABILITY AND EXPOSURE TO CRIME: APPLYING RISK TERRAIN MODELING TO THE STUDY OF ASSAULT IN CHICAGO L. W. Kennedy J. M. Caplan E. L. Piza H. Buccine- Schraeder Full Article: Kennedy, L. W., Caplan,

More information

Two-Way Independent ANOVA

Two-Way Independent ANOVA Two-Way Independent ANOVA Analysis of Variance (ANOVA) a common and robust statistical test that you can use to compare the mean scores collected from different conditions or groups in an experiment. There

More information

CHAPTER 6 HUMAN BEHAVIOR UNDERSTANDING MODEL

CHAPTER 6 HUMAN BEHAVIOR UNDERSTANDING MODEL 127 CHAPTER 6 HUMAN BEHAVIOR UNDERSTANDING MODEL 6.1 INTRODUCTION Analyzing the human behavior in video sequences is an active field of research for the past few years. The vital applications of this field

More information

Statistics is the science of collecting, organizing, presenting, analyzing, and interpreting data to assist in making effective decisions

Statistics is the science of collecting, organizing, presenting, analyzing, and interpreting data to assist in making effective decisions Readings: OpenStax Textbook - Chapters 1 5 (online) Appendix D & E (online) Plous - Chapters 1, 5, 6, 13 (online) Introductory comments Describe how familiarity with statistical methods can - be associated

More information

Case Studies of Signed Networks

Case Studies of Signed Networks Case Studies of Signed Networks Christopher Wang December 10, 2014 Abstract Many studies on signed social networks focus on predicting the different relationships between users. However this prediction

More information

Bayesian Logistic Regression Modelling via Markov Chain Monte Carlo Algorithm

Bayesian Logistic Regression Modelling via Markov Chain Monte Carlo Algorithm Journal of Social and Development Sciences Vol. 4, No. 4, pp. 93-97, Apr 203 (ISSN 222-52) Bayesian Logistic Regression Modelling via Markov Chain Monte Carlo Algorithm Henry De-Graft Acquah University

More information

Chapter 2 Interactions Between Socioeconomic Status and Components of Variation in Cognitive Ability

Chapter 2 Interactions Between Socioeconomic Status and Components of Variation in Cognitive Ability Chapter 2 Interactions Between Socioeconomic Status and Components of Variation in Cognitive Ability Eric Turkheimer and Erin E. Horn In 3, our lab published a paper demonstrating that the heritability

More information

Regression Including the Interaction Between Quantitative Variables

Regression Including the Interaction Between Quantitative Variables Regression Including the Interaction Between Quantitative Variables The purpose of the study was to examine the inter-relationships among social skills, the complexity of the social situation, and performance

More information

SBIRT IOWA. Iowa Army National Guard THE IOWA CONSORTIUM FOR SUBSTANCE ABUSE RESEARCH AND EVALUATION. Iowa Army National Guard

SBIRT IOWA. Iowa Army National Guard THE IOWA CONSORTIUM FOR SUBSTANCE ABUSE RESEARCH AND EVALUATION. Iowa Army National Guard SBIRT IOWA Iowa Army National Guard THE IOWA CONSORTIUM FOR SUBSTANCE ABUSE RESEARCH AND EVALUATION Iowa Army National Guard Biannual Report October 2016 With Funds Provided By: Iowa Department of Public

More information

Design and Analysis Plan Quantitative Synthesis of Federally-Funded Teen Pregnancy Prevention Programs HHS Contract #HHSP I 5/2/2016

Design and Analysis Plan Quantitative Synthesis of Federally-Funded Teen Pregnancy Prevention Programs HHS Contract #HHSP I 5/2/2016 Design and Analysis Plan Quantitative Synthesis of Federally-Funded Teen Pregnancy Prevention Programs HHS Contract #HHSP233201500069I 5/2/2016 Overview The goal of the meta-analysis is to assess the effects

More information

Meta-Analysis and Subgroups

Meta-Analysis and Subgroups Prev Sci (2013) 14:134 143 DOI 10.1007/s11121-013-0377-7 Meta-Analysis and Subgroups Michael Borenstein & Julian P. T. Higgins Published online: 13 March 2013 # Society for Prevention Research 2013 Abstract

More information

Assessing Studies Based on Multiple Regression. Chapter 7. Michael Ash CPPA

Assessing Studies Based on Multiple Regression. Chapter 7. Michael Ash CPPA Assessing Studies Based on Multiple Regression Chapter 7 Michael Ash CPPA Assessing Regression Studies p.1/20 Course notes Last time External Validity Internal Validity Omitted Variable Bias Misspecified

More information

MCAS Equating Research Report: An Investigation of FCIP-1, FCIP-2, and Stocking and. Lord Equating Methods 1,2

MCAS Equating Research Report: An Investigation of FCIP-1, FCIP-2, and Stocking and. Lord Equating Methods 1,2 MCAS Equating Research Report: An Investigation of FCIP-1, FCIP-2, and Stocking and Lord Equating Methods 1,2 Lisa A. Keller, Ronald K. Hambleton, Pauline Parker, Jenna Copella University of Massachusetts

More information

Chapter 11. Experimental Design: One-Way Independent Samples Design

Chapter 11. Experimental Design: One-Way Independent Samples Design 11-1 Chapter 11. Experimental Design: One-Way Independent Samples Design Advantages and Limitations Comparing Two Groups Comparing t Test to ANOVA Independent Samples t Test Independent Samples ANOVA Comparing

More information

9 research designs likely for PSYC 2100

9 research designs likely for PSYC 2100 9 research designs likely for PSYC 2100 1) 1 factor, 2 levels, 1 group (one group gets both treatment levels) related samples t-test (compare means of 2 levels only) 2) 1 factor, 2 levels, 2 groups (one

More information

Readings: Textbook readings: OpenStax - Chapters 1 13 (emphasis on Chapter 12) Online readings: Appendix D, E & F

Readings: Textbook readings: OpenStax - Chapters 1 13 (emphasis on Chapter 12) Online readings: Appendix D, E & F Readings: Textbook readings: OpenStax - Chapters 1 13 (emphasis on Chapter 12) Online readings: Appendix D, E & F Plous Chapters 17 & 18 Chapter 17: Social Influences Chapter 18: Group Judgments and Decisions

More information

Two-Way Independent Samples ANOVA with SPSS

Two-Way Independent Samples ANOVA with SPSS Two-Way Independent Samples ANOVA with SPSS Obtain the file ANOVA.SAV from my SPSS Data page. The data are those that appear in Table 17-3 of Howell s Fundamental statistics for the behavioral sciences

More information

Doing Quantitative Research 26E02900, 6 ECTS Lecture 6: Structural Equations Modeling. Olli-Pekka Kauppila Daria Kautto

Doing Quantitative Research 26E02900, 6 ECTS Lecture 6: Structural Equations Modeling. Olli-Pekka Kauppila Daria Kautto Doing Quantitative Research 26E02900, 6 ECTS Lecture 6: Structural Equations Modeling Olli-Pekka Kauppila Daria Kautto Session VI, September 20 2017 Learning objectives 1. Get familiar with the basic idea

More information

MS&E 226: Small Data

MS&E 226: Small Data MS&E 226: Small Data Lecture 10: Introduction to inference (v2) Ramesh Johari ramesh.johari@stanford.edu 1 / 17 What is inference? 2 / 17 Where did our data come from? Recall our sample is: Y, the vector

More information

Student name: SOCI 420 Advanced Methods of Social Research Fall 2017

Student name: SOCI 420 Advanced Methods of Social Research Fall 2017 SOCI 420 Advanced Methods of Social Research Fall 2017 EXAM 1 RUBRIC Instructor: Ernesto F. L. Amaral, Assistant Professor, Department of Sociology Date: October 12, 2017 (Thursday) Section 904: 2:20 3:35pm

More information

Lessons in biostatistics

Lessons in biostatistics Lessons in biostatistics The test of independence Mary L. McHugh Department of Nursing, School of Health and Human Services, National University, Aero Court, San Diego, California, USA Corresponding author:

More information

Results & Statistics: Description and Correlation. I. Scales of Measurement A Review

Results & Statistics: Description and Correlation. I. Scales of Measurement A Review Results & Statistics: Description and Correlation The description and presentation of results involves a number of topics. These include scales of measurement, descriptive statistics used to summarize

More information

Minimizing Uncertainty in Property Casualty Loss Reserve Estimates Chris G. Gross, ACAS, MAAA

Minimizing Uncertainty in Property Casualty Loss Reserve Estimates Chris G. Gross, ACAS, MAAA Minimizing Uncertainty in Property Casualty Loss Reserve Estimates Chris G. Gross, ACAS, MAAA The uncertain nature of property casualty loss reserves Property Casualty loss reserves are inherently uncertain.

More information

Regression CHAPTER SIXTEEN NOTE TO INSTRUCTORS OUTLINE OF RESOURCES

Regression CHAPTER SIXTEEN NOTE TO INSTRUCTORS OUTLINE OF RESOURCES CHAPTER SIXTEEN Regression NOTE TO INSTRUCTORS This chapter includes a number of complex concepts that may seem intimidating to students. Encourage students to focus on the big picture through some of

More information

Simple Linear Regression the model, estimation and testing

Simple Linear Regression the model, estimation and testing Simple Linear Regression the model, estimation and testing Lecture No. 05 Example 1 A production manager has compared the dexterity test scores of five assembly-line employees with their hourly productivity.

More information

Iowa Army National Guard Biannual Report April 2016

Iowa Army National Guard Biannual Report April 2016 SBIRT IOWA Iowa Army National Guard THE IOWA CONSORTIUM FOR SUBSTANCE ABUSE RESEARCH AND EVALUATION Iowa Army National Guard Biannual Report April 2016 With Funds Provided By: Iowa Department of Public

More information

TERRORIST WATCHER: AN INTERACTIVE WEB- BASED VISUAL ANALYTICAL TOOL OF TERRORIST S PERSONAL CHARACTERISTICS

TERRORIST WATCHER: AN INTERACTIVE WEB- BASED VISUAL ANALYTICAL TOOL OF TERRORIST S PERSONAL CHARACTERISTICS TERRORIST WATCHER: AN INTERACTIVE WEB- BASED VISUAL ANALYTICAL TOOL OF TERRORIST S PERSONAL CHARACTERISTICS Samah Mansoour School of Computing and Information Systems, Grand Valley State University, Michigan,

More information

Exploring the Impact of Missing Data in Multiple Regression

Exploring the Impact of Missing Data in Multiple Regression Exploring the Impact of Missing Data in Multiple Regression Michael G Kenward London School of Hygiene and Tropical Medicine 28th May 2015 1. Introduction In this note we are concerned with the conduct

More information

Missing Data and Institutional Research

Missing Data and Institutional Research A version of this paper appears in Umbach, Paul D. (Ed.) (2005). Survey research. Emerging issues. New directions for institutional research #127. (Chapter 3, pp. 33-50). San Francisco: Jossey-Bass. Missing

More information

Stepwise Knowledge Acquisition in a Fuzzy Knowledge Representation Framework

Stepwise Knowledge Acquisition in a Fuzzy Knowledge Representation Framework Stepwise Knowledge Acquisition in a Fuzzy Knowledge Representation Framework Thomas E. Rothenfluh 1, Karl Bögl 2, and Klaus-Peter Adlassnig 2 1 Department of Psychology University of Zurich, Zürichbergstraße

More information

Content. Basic Statistics and Data Analysis for Health Researchers from Foreign Countries. Research question. Example Newly diagnosed Type 2 Diabetes

Content. Basic Statistics and Data Analysis for Health Researchers from Foreign Countries. Research question. Example Newly diagnosed Type 2 Diabetes Content Quantifying association between continuous variables. Basic Statistics and Data Analysis for Health Researchers from Foreign Countries Volkert Siersma siersma@sund.ku.dk The Research Unit for General

More information

POLS 5377 Scope & Method of Political Science. Correlation within SPSS. Key Questions: How to compute and interpret the following measures in SPSS

POLS 5377 Scope & Method of Political Science. Correlation within SPSS. Key Questions: How to compute and interpret the following measures in SPSS POLS 5377 Scope & Method of Political Science Week 15 Measure of Association - 2 Correlation within SPSS 2 Key Questions: How to compute and interpret the following measures in SPSS Ordinal Variable Gamma

More information

Investigating the robustness of the nonparametric Levene test with more than two groups

Investigating the robustness of the nonparametric Levene test with more than two groups Psicológica (2014), 35, 361-383. Investigating the robustness of the nonparametric Levene test with more than two groups David W. Nordstokke * and S. Mitchell Colp University of Calgary, Canada Testing

More information

Basic concepts and principles of classical test theory

Basic concepts and principles of classical test theory Basic concepts and principles of classical test theory Jan-Eric Gustafsson What is measurement? Assignment of numbers to aspects of individuals according to some rule. The aspect which is measured must

More information