Using Ideal Observers in Higher-order Human Category Learning

Size: px
Start display at page:

Download "Using Ideal Observers in Higher-order Human Category Learning"

Transcription

1 Usig Ideal Observers i Higher-order Huma ategory Learig Aiket Kittur (kittur@ucla.edu) Keith J. Holyoak (holyoak@lifesci.ucla.edu) Departmet of Psychology, Uiversity of aliforia, Los Ageles A Joh E. Hummel (jehummel@cyrus.psych.uiuc.edu) Departmet of Psychology, Uiversity of Illiois at Urbaa-hampage, Urbaa IL 6820 Abstract Ideal observer models have prove useful i ivestigatig assumptios about huma iformatio processig i a variety of perceptual tasks. However, these models have ot bee applied i the area of higher-order category learig. We describe a simple Bayesia ideal observer ad apply it to empirical data o category learig. We describe a experimet i which we foud that acquisitio of family resemblace categories was drastically impaired if the categories were defied by relatios betwee features rather tha by the features themselves. A ideal observer was used to test whether this effect could be accouted for by iheret iformatio differeces betwee the coditios. A compariso of participats performace to the model foud a sigificat differece i efficiecy of learig eve after accoutig for iformatio differeces betwee coditios. This aalysis illustrates how ideal observer methods ca provide useful tools for aalyzig higher-order category learig. Keywords: categorizatio, category learig, relatios, features, ideal observer Itroductio A ideal observer model makes optimal use of a set of give iformatio i performig a task. Such models have traditioally prove useful i ivestigatig huma iformatio use i various perceptual tasks by providig a upper boud or bechmark by which to measure performace. If a huma ca perform at the same level as (or better tha) the ideal model, the we kow that the huma is makig use of all of the available iformatio i the situatio (or, i the case of humas outperformig the ideal, more iformatio tha was available to the model). If humas uderperform the ideal model, the differece ca ofte highlight specific costraits that limit huma iformatio processig. The degree to which huma performace approaches that of a ideal observer ca provide a measure of processig efficiecy. Ideal observers have most commoly bee applied to uderstadig huma low-level visual tasks ivolvig detectio ad discrimiatio (see Geisler, 2003), though they have also bee applied to tasks such as readig (Legge, Klitz, & Tja, 997), object recogitio (Liu, Kill, & Kerste, 995), ad reachig (Trommershäuser, Gepshtei, Maloey, Lady, & Baks, 2004). However, few studies have applied ideal observer methods to higher-order cogitive tasks, at least i part because of the difficulty of specifyig exactly what is ideal. Istead, most studies of huma category learig compare coditios agaist each other ad assume that differeces i performace capture theoretically-cetral differeces betwee coditios. However, there may be differeces betwee coditios that are ot relevat to the variable beig measured (e.g., oise). Ideal observers ca provide a theoretical upper boud o huma performace (give a set of assumptios), ad ca be used to cotrol for some of these extraeous variables. I this paper we describe a simple method for creatig a ideal observer model that takes as iput features, ad relatios betwee features, of the sort commoly used i category learig studies with artificial stimuli. The ideal observer assumes that the experimeter (but ot ecessarily the learer) has full kowledge of the geeratig model used to costruct stimuli based o these features ad relatios. This assumptio is typically met i experimetal category learig paradigms i which artificial stimuli are used. We will first describe the model ad the apply it by simulatig performace i a actual category learig experimet. The Model The model uses a Bayesia framework to assig stimuli to categories ad to lear from labeled feedback. We use a versio of a aïve Bayesia classifier, oe of the simplest probabilistic classifiers, which is optimal whe all iput features are idepedet (ad ca eve be optimal i certai less restricted circumstaces; see Domigos & Pazzai, 997). The aïve Bayesia classifier makes the assumptio that all features of a give category are geerated idepedetly, that is:, F,..., F ) = ) F )... F ) () for class variable (which represets all possible categories) ad feature variables through. Applyig Bayes rule results i the followig equatio: F,..., F ) = ) i= [ ) F ) i= i F )] The deomiator i Equatio 2 is a ormalizatio costat that is idetical for all categories ad thus ofte igored for simplicity (though implemeted i the model). With two equally-probable categories (as is most commo i category learig paradigms) ) is also costat (.5), ad thus the mai determiat of classificatio is F i ). This probability is calculated i the followig maer: i (2)

2 where F i F i + α Fi Fi ) = (3) + α is the umber of items with feature F i i category, c is the total umber of items i category, ad α F i ad α are uiform priors. Equatio 3 ca be iterpreted as updatig a uiform prior with ew iformatio, with the prior evetually overwhelmed as more features are observed. The classifier ca be exteded to reflect uderlyig depedecies betwee features that are ot idepedetly geerated. This refiemet ca ofte be useful i categorizatio studies whe oe feature costrais the values of other features. Whe these depedecies are kow, they ca be icorporated ito the model by retaiig the relevat coditioal probabilities. For example, Equatio 4 is a toy model with two features i which feature 2 is depedet o feature (the ormalizatio costat is left out for simplicity): p F, F ) = ) F ) F, ) (4) ( 2 2 F Equatio 3 ca also be exteded for features that are depedet o other features, becomig: F i, F + α j Fi Fi, F j ) = (5) + α where F i is depedet o feature F j. The algorithm for category learig operates as follows. First, a ew example is preseted to the model without category label iformatio. The probability of its beig i each category is calculated based o previously-observed labeled examples, ad the resultig probabilities are used to assig a predicted category to the example 2. The typical way to classify a ew example is to choose the category with the maximum probability of geeratig the example (Geisler, 2003). Oce category assigmet is complete the example is placed ito the observed set alog with its category label. This step simulates the effect of feedback, with the ew example ow affectig future classificatio judgmets. Order of presetatio is importat: like the participats, the model s predictios ca oly be based o previously see exemplars. There are may ways to compare the model s performace with humas. Oe possibility is to use a metric based o the umber of correct ad icorrect trials, which is perhaps the closest aalog to how ideal observers have bee used i recet studies (Geisler, 2003). However, may More specifically, the α Fi s describe the parameters of a Dirichlet prior i which all values are set to, with their sum beig α c. 2 Oe issue with the algorithm is how to get it started. Although there are a umber of justifiable methods, here we start the model with the smallest umber of examples for which there is oe example for each category. category learig paradigms focus o differeces i learig rates, with a commo metric beig the umber of trials eeded to reach a certai performace criterio. Viewed i terms of statistical samplig (e.g., the umber of samples eeded to lear a certai distributio), this metric provides a atural compariso of huma ad model learig. Specifically, we ca defie samplig efficiecy as the ratio of the umber of trials the model eeded to lear to criterio to the umber of trials a huma eeded (see Scholkopf & Smola, 2002; Stakiewitz, 2003): ttc mod (6) ttc par where ttc mod is the trials to criterio eeded for the model ad ttc par the trials eeded for the participat. The closer huma performace comes to the ideal, the higher the efficiecy. I may ways the ideal observer described here is similar to Aderso s (99) ratioal theory of categorizatio. However, Aderso focuses o determiig the optimal categorizatio give a geeral eviromet i pursuit of a descriptive theory of huma categorizatio. I cotrast, our ideal observer simply aims to be ormative i a specific eviromet for which the structure ad geeratig model is kow, ad to provide a bechmark or upper boud o huma performace. Thus may of the goals ad assumptios of Aderso s model are very differet from the ideal observer described here. For example, sice we kow ad capture the depedecies betwee features, we do ot make the simplifyig assumptio that all features are coditioally idepedet. Droppig this assumptio is ecessary i order to maitai optimality for the types of geeratig models commoly used i higher-order category learig, where features are ofte costraied by the values of other relatios or features. Also, istead of predictig a usee feature (such as the category label) through chaied iferece, we focus o the simpler task of predictig a category class give a set of features. This simpler goal allows us to avoid usig weighted category averages ad oly requires computatio of the maximum likelihood category. Fially, we avoid the eed for a empirically-determied variable goverig the probability of creatig a ew category (Aderso s couplig probability ). We will ow apply this ideal observer i order to model learig i a categorizatio experimet i which the differet coditios may have differet types ad amouts of iformatio associated with them. The Experimet A fudametal shift i the uderstadig of categorizatio resulted from the family resemblace view of categories, which argued that may categories have a graded structure based o shared features (Medi & Schaffer, 978; Rosch, 976; Rosch & Mervis, 975; Wittgestei, 953). The family resemblace view has had great success accoutig for peoples learig ad geeralizatio of categories that ca be represeted as simple lists of features. Such

3 categories ca be leared implicitly ad automatically, with feature-category associatios ot ecessarily available to coscious verbalizatio (Ashby, Maddox, & Bohil, 2002; Ashby & Waldro, 999). However, much of huma coceptual kowledge is composed of categories that caot be represeted as simple features (Barsalou, 983; Keil, 989; Murphy & Medi, 985; Rips, 989; Ross & Spaldig, 994). Rather, may cocepts are based o the relatioships betwee thigs rather tha the literal features of the thigs themselves. For example, a barrier is a relatioal cocept that ca be as cocrete as a wall or moat or as abstract as a lack of moey or the color of oe s ski. Relatioal cocepts aboud i everyday life, with examples icludig social uderstadig (a love triagle), law (breach of cotract), religio (atoemet for sis), sciece (coservatio of eergy), as well as basic perceptio (recogizig arragemets of objects as scees) (e.g., Geter & Kurtz, 2005, Holyoak & Thagard, 995). Although relatioal cocepts are fudametal to huma itelligece, our uderstadig of how we lear them is poor compared to our uderstadig of feature-based categories. A reasoable ad parsimoious hypothesis is that relatioal categories act just like feature-based categories with the features replaced by relatios that is, cocept learig may be a sigle uified process that ca take either features or relatios as iput. This view predicts that relatioal categories should show the same kid of family resemblace structure evideced by feature-based categories, thus geeralizig what we have leared about category learig from feature-based to relatioal categories. However, there is evidece that relatios ad features may be psychologically distict. For example, Medi, Goldstoe, ad Geter (990, 993) demostrated strog empirical differeces betwee relatioal ad feature-based similarity, suggestig that relatios ad features may rely o separate, competig processes for assessig similarity. osistet with these fidigs, some researchers have argued that feature lists are fudametally iadequate to represet relatioal cocepts, ad that such cocepts must istead be metally represeted as relatioal structures such as schemas or theories (Geter, 983; Hollad, Holyoak, Nibett, & Thagard, 986; Hummel & Holyoak, 2003; Keil, 989; Murphy & Medi, 985). I such accouts, learig a relatioal category is more aki to iducig a schema tha to learig a list of diagostic features. Most accouts of schema iductio assume that a shared, determiistic cohesive elemet is ecessary to create the schema i the first place (Hummel & Holyoak, 2003; Kuehe et al., 2000). We coducted a experimet to test whether relatioal ad feature-based categories were leared i similar ways 3. Specifically, we hypothesized that relatioal categories i which o sigle defiig elemet existed as is the case i family resemblace categories would prove drastically more difficult to lear tha feature-based categories with a 3 The experimet described here is based o pilot data reported i Kittur, Hummel, ad Holyoak (2004), which describes i more detail the methods used ad additioal measures collected. idetical family resemblace structure. Whereas learig family resemblace categories based o simple features may be doe implicitly through trackig ad averagig the features of the exemplars of each category, learig relatioal categories will be much more difficult because the same feature(s) may be associated with multiple categories, depedig o the relatios ivolved. To test this hypothesis we used a 2x2 betwee-subjects desig, i which categories either had a sigle dimesio perfectly predictive of category membership (determiistic) or had a family resemblace structure i which three out of four dimesios were characteristic of the category but o sigle dimesio was perfectly predictive. Dimesios were defied either by idividual feature values or by the relatios betwee features. We predicted a iteractio betwee category structure ad type, such that the relatioal family resemblace coditio would be much more difficult to lear tha ay of the other three coditios. Method Subjects. 96 Uiversity of aliforia, Los Ageles udergraduates participated for partial fulfillmet of course requiremets. Figure. Examples of family resemblace categories. Determiistic relatioal categories were formed by removig oe exemplar from each category. The table depicts the dimesios categories were defied o, as well as the value of each exemplar o the dimesios (filled = value, empty = value 2). For example, i relatioal category exemplar i, the octago (O) is bigger, darker, above, ad behid the square (S). Show are oly a small subset of all istatiatios of the four exemplar types for a category.

4 Stimuli ad Procedure. All stimuli were composed of a octago ad a square set i a fixed backgroud resemblig a computer chip. Either the relatios betwee the two shapes (relatioal coditio) or the idividual features of each shape (feature-based coditio) determied category membership of each exemplar. Relatioal categories were defied by whether the octago was ) larger, 2) darker, 3) vertically above, ad 4) i frot of the square (see Figure ). Feature-based categories were defied by idividual absolute feature values: ) size of the octago, 2) color of the octago, 3) size of the square, ad 4) color of the square. rossed with the feature-based ad relatioal coditios was the structure of each category. I the family resemblace coditio, each category member had three out of four dimesios cosistet with its category ad oe icosistet dimesio. I the determiistic coditio oe dimesio was perfectly diagostic across all exemplars. This desig yielded four coditios to which participats were radomly assiged: relatioal family resemblace or determiistic (R-FR or R-D) ad feature-based family resemblace or determiistic (F-FR or F-D). O each trial of the acquisitio phase, a participat viewed oe exemplar, categorized it as a math or graphics chip, ad received accuracy feedback. Acquisitio cotiued util the participat reached criterio (>88% correct for two cosecutive blocks 4 ). Behavioral Results The relatioal family resemblace coditio proved much more difficult to lear tha the other three coditios: 22% of participats i the relatioal family resemblace coditio did ot lear to criterio withi 600 trials (o participats i ay other coditios failed to lear). All results make the extremely coservative assumptio that participats who failed to lear would have succeeded o trial 60. The mea umber of trials to criterio for each coditio is show i Figure 2a. There were mai effects of both category type (relatios vs. features, F(, 95) = 4.7, p =.032, ad category structure (family resemblace vs. determiistic, F(, 95) = 9.83, p =.002; importatly, there was also a sigificat iteractio of category type ad structure, F(,95) = 6.4, p =.05, due to extremely impaired acquisitio whe the category was defied by relatios ad had a family resemblace structure. Ideal Observer Aalysis Oe explaatio of these results is that relatios ad features are represeted ad processed differetly i the brai, ad that relatioal categories may ot have access to the machiery that is used to lear feature-based family resemblace categories. However, aother explaatio could be that the selective impairmet of the relatioal family resem- 4 This criterio was chose so that simple feature-trackig strategies (e.g., memorizig the associatios of sigle features with categories) would lead to sub-criterio performace. Trials to riterio (Huma) Trials to riterio (Model) Efficiecy (Model/Huma) Featural Featural Featural Det FR (a) Relatioal Relatioal Relatioal (b) Figure 2. (a) Mea trials to criterio take by participats to lear the categories. Det=Determiistic, FR=Family Resemblace. (b) Mea trials to criterio for the model to lear the categories give the same stimuli i the same order as each idividual participat. (c) Efficiecy of huma performace compared to model performace measured by the ratio of the umber of trials eeded by the model to lear to criterio to the umber of trials eeded by huma learers. (Note that efficiecy is calculated o a per-subject basis, ad so caot be determied from paels (a) ad (b) aloe.) blace coditio is istead due to a differece i the amout of available diagostic iformatio. I other words, are people worse oly because some coditios are iheretly more difficult to lear due to lack of iformatio? To aswer this questio we adapted the ideal observer model described earlier to the curret experimetal task. The features ad relatios available to participats were coded as discrete values o separate dimesios ad used as iputs to the model. For example, the model received as separate iputs the size of the square, the size of the octago, ad the relatio of which was bigger. The same iformatio was available to participats, who could use iformatio about either the features or the relatios o each trial. However, the relatioal ad featural iformatio o a dimesio were ot idepedet: i the example above, if the relatio was octago bigger tha square, the kowig the size of the octago provides iformatio about the size of the square (which must be smaller; see Figure 3). To accout for this depedecy, relatios were modeled as idepedet (c)

5 Figure 3. Relatios ad features ivolved i category geeratio. Arrows depict depedecies (i.e., costraits) betwee relatios ad features. iputs, whereas feature iputs were coditioal o their respective relatio. 5 For the ideal observer described here to be truly optimal, the geeratig model must meet certai assumptios. First, the distributio of category members must be sampled from idepedet multiomial distributios with Dirichletdistributed parameters. This assumptio holds true: category members were geerated by samplig from idepedet multiomial distributios with a equal likelihood of each member appearig. Secod, all depedecies that arise i geeratig feature values for each category member must be captured i the coditioal probabilities (i.e., relatios costraiig features). This assumptio is also valid: the depedecies show i Figure 3 reflect how the exemplars were geerated. The ideal observer model was ru o each participat s data, ad the umber of trials ecessary to lear to criterio was measured 6 (see Figure 2b). We the computed each participat s efficiecy accordig to Equatio 6. These efficiecies are depicted i Figure 2c. Huma performace as measured by statistical efficiecy was much worse i the relatioal family resemblace coditio tha i ay other coditio. A ANOVA was performed o the efficiecy measure followig a log trasformatio to ormalize the variaces. The results demostrated that the critical iteractio was sigificat, F(, 95) = 3.93, p <.05. Sice efficiecy takes ito accout differeces i model as well as huma performace, fidig a iteractio o this measure idicates that the huma learig rates for these coditios were more differet tha would expected give the iheret difficulty of the coditios. That is, iheret iformatioal differeces betwee the coditios were isufficiet to accout for the disparities i huma performace. Discussio The behavioral results revealed a clear impairmet i acquisitio for relatioal categories defied by a family resemblace structure, as compared to categories based o features, which are leared quickly whether they had family resemblace or determiistic structure. Relatioal catego- 5 A atural questio is: should the features be defiig i the featural coditios, rather tha the relatios? No chage is eeded i the model because i the featural coditios the dimesios o which the features were cosidered depedet (relative size ad shade) had the same relatioal value for both categories. Thus the features become effectively idepedet. 6 A more statistically accurate phrasig of this would read: the umber of samples eeded to lear the distributio to a certai degree of accuracy. ries with determiistic structure were leared as quickly as determiistic feature-based categories, suggestig that the effect is ot merely due to the relatioal ature of the task. This iteractio is icosistet with the hypothesis that relatioal categories are leared i the same way as featurebased categories. A ideal observer aalysis was used to determie whether this impaired learig might be due to iheret iformatioal differeces betwee coditios. By comparig the efficiecy of huma performace to that of the ideal model, we were able to show that objective differeces i difficulty betwee coditios did ot accout for the experimetal data. Rather, it appears that relatios ad features are represeted or processed differetly i huma category learig. Idetifyig exactly how relatios ad features differ is a importat subject for future research. Oe potetially useful approach is to determie what chages to the ideal observer could make it more closely match huma data. For example, what happes whe the model does ot have perfect memory, or caot perfectly update its prior? Or whe its workig memory is impaired so it caot atted to all relatios ad features at oce? Observig how the model degrades as additioal costraits are added could provide valuable isights ito huma iformatio processig. Alteratively, it is possible that o processig-related chages i the ideal observer will capture the dissociatio i the huma data. Istead, it may be ecessary to take ito accout the represetatioal differece betwee simple features ad relatioal predicates. It remais a ope questio how to icorporate structured predicates ito a Bayesia framework; extat aalyses of categorizatio usig Bayesia iferece treat relatios as correlatios or ustructured features rather tha as structured predicates (e.g., Kemp et al., 2004). Ideed, oe possible explaatio of our results may be that the likelihood updatig mechaism at work i featural categorizatio may ot be used for relatioal categorizatio, resultig i impairmet of relatioal family resemblace learig. At first glace the preset results appear couterituitive: relatioal category learig is severely impaired if o elemets are costat across all exemplars, yet people seem able to coceptualize family resemblace relatioal categories, such as Wittgestei s (953) classic example of game. This paradox highlights the eed for additioal empirical studies. Oe approach to explorig this seemig icosistecy may be to examie prior kowledge ad experiece. While a sigle costat elemet may be ecessary to lear ovel relatioal categories, whe prior kowledge ad experiece are brought ito play this critical eed may be reduced. It is possible that a coheret theory that explais a relatioal family resemblace structure might make learig easier (Rehder & Hastie, 2004; Rehder & Ross, 200). I additio, repeated experiece with the relevat relatios may lead to low-level chukig of a stimulus, as i chess experts memory for board positios. Thus both higher-order causal explaatios ad lower-order experiece may facilitate relatioal learig.

6 I summary, the dissociatio betwee feature-based category learig, which is robust to family-resemblace structure, ad relatio-based category learig, which is ot, suggests that curret feature-based models of category learig may have limited applicability to relatioal categories. The difficulty for such models is ot oly that feature lists are iadequate to represet relatios, but that the two kids of categories are processed differetly as well. The preset study also demostrates how ideal observer methods ca be applied i higher-order category learig. Here we used a ideal observer to provide a objective measure of the ease of learig i each coditio. The model is easy to implemet i category learig studies with discrete stimuli for which the geeratig model is kow. We believe that the ideal-observer approach ca have geeral applicability for studies of category learig i which differet learig coditios may have differet iformatioal cotet. Ackowledgmets Preparatio of this paper was supported by NSF grat SES to KH. Special thaks to Hogjig Lu, Zili Liu, ad Barbara Spellma for helpful commets ad discussio. Refereces Aderso, J. R. (99). The adaptive ature of huma categorizatio. Psychological Review, 98, Ashby, F. G., Maddox, W. T., & Bohil,. J. (2002). Observatioal versus feedback traiig i rule-based ad iformatio-itegratio category learig. Memory & ogitio, 30, Ashby, F. G., & Waldro, E. M. (999). O the ature of implicit categorizatio. Psychoomic Bulleti & Review, 6, Barsalou, L. W. (983). Ad hoc categories. Memory & ogitio,, Domigos, P., & Pazzai, M. J. (997). O the optimality of the simple Bayesia classifier uder zero-oe loss. Machie Learig, 29, Geisler, W. S. (2003). Ideal Observer aalysis. I L. halupa, ad J. Werer (Eds.), The visual euroscieces (pp ). Bosto: MIT Press. Geter, D., & Kurtz, K. J. (2005). Learig ad usig relatioal categories. I W. K. Ah, Goldstoe, R.L., Love, B.., Markma, A.B., & Wolff, P.W. (Ed.), ategorizatio iside ad outside the lab. Washigto, D..: APA. Geter, D. (983). Structure-mappig: A theoretical framework for aalogy. ogitive Sciece, 7, Hollad, J. H., Holyoak, K. J., Nisbett, R. E., & Thagard, P. R. (986). Iductio: Processes of Iferece, Learig, ad Discovery. ambridge, MA, US: The MIT Press. Holyoak, K. J., & Thagard, P. (995). Metal leaps: Aalogy i creative thought. ambridge, MA: The MIT Press. Hummel, J. E., & Holyoak, K. J. (2003). A symboliccoectioist theory of relatioal iferece ad geeralizatio. Psychological Review, 0, Keil, F.. (989). ocepts, kids, ad cogitive developmet. ambridge: The MIT Press. Kemp,., Griffiths, T. L., & Teebaum, J. B. (2004). Discoverig latet classes i relatioal data. MIT AI Memo Kuehe, S., Forbus, K. D., Geter, D., & Qui, B. (2000). SEQL: ategory learig as progressive abstractio usig structure mappig. Proceedigs of the Twety-secod Aual oferece of the ogitive Sciece Society. Legge, G. E., Klitz, T. S., & Tja, B. S. (997). Mr. hips: A ideal-observer model of readig. Psychological Review, 04, Liu, Z., Kill, D.., & Kerste, D. (995). Object classificatio for huma ad ideal observers. Visio Research, 35, Medi, D. L., Goldstoe, R. L., & Geter, D. (990). Similarity ivolvig attributes ad relatios: Judgmets of similarity ad differece are ot iverses. Psychological Sciece,, Medi, D. L., Goldstoe, R. L., & Geter, D. (993). Respects for similarity. Psychological Review, 00, Medi, D. L., & Schaffer, M. M. (978). otext theory of classificatio learig. Psychological Review, 85, Murphy, G. L. (2002). The big book of cocepts. ambridge: The MIT Press Murphy, G. L., & Medi, D. L. (985). The role of theories i coceptual coherece. Psychological Review, 92, Rehder, B., & Hastie, R. (2004). ategory coherece ad category-based property iductio. ogitio, 9, Rehder, B., & Ross, B. H. (200). Abstract coheret categories. Joural of Experimetal Psychology: Learig, Memory, & ogitio, 27, Rips, L. J. (989). Similarity, typicality, ad categorizatio. I S. O. Vosiadou, Adrew (Ed.), Similarity ad aalogical reasoig. New York: ambridge Uiversity Press. Rosch, E., & Mervis,. B. (975). Family resemblaces: Studies i the iteral structure of categories. ogitive Psychology, 7, Rosch, E., Simpso,., & Miller, R. S. (976). Structural bases of typicality effects. Joural of Experimetal Psychology: Huma Perceptio & Performace, 2, Ross, B. H., & Spaldig, T. L. (994). ocepts ad categories. I R. J. Sterberg (Ed.), Thikig ad problem solvig. Hadbook of perceptio ad cogitio (2d ed.) (pp. 9-48). Sa Diego, A: Academic Press, Ic. Scholkopf, B., & Smola, A. J. (2002). Learig with Kerels: MIT Press. Stakiewicz, B.J., Legge, G.E., Masfield, J.S., & Schlicht, E.J. (i press). Lost i virtual space: Studies i huma ad ideal spatial avigatio. Joural of Experimetal Psychology: Huma Perceptio ad Performace. Wittgestei, L. (953). Philosophical ivestigatios. New York: Macmilla.

Objectives. Sampling Distributions. Overview. Learning Objectives. Statistical Inference. Distribution of Sample Mean. Central Limit Theorem

Objectives. Sampling Distributions. Overview. Learning Objectives. Statistical Inference. Distribution of Sample Mean. Central Limit Theorem Objectives Samplig Distributios Cetral Limit Theorem Ivestigate the variability i sample statistics from sample to sample Fid measures of cetral tedecy for distributio of sample statistics Fid measures

More information

Caribbean Examinations Council Secondary Education Certificate School Based Assessment Additional Math Project

Caribbean Examinations Council Secondary Education Certificate School Based Assessment Additional Math Project Caribbea Examiatios Coucil Secodary Educatio Certificate School Based Assessmet Additioal Math Project Does good physical health ad fitess, as idicated by Body Mass Idex, affect the academic performace

More information

Statistics 11 Lecture 18 Sampling Distributions (Chapter 6-2, 6-3) 1. Definitions again

Statistics 11 Lecture 18 Sampling Distributions (Chapter 6-2, 6-3) 1. Definitions again Statistics Lecture 8 Samplig Distributios (Chapter 6-, 6-3). Defiitios agai Review the defiitios of POPULATION, SAMPLE, PARAMETER ad STATISTIC. STATISTICAL INFERENCE: a situatio where the populatio parameters

More information

23.3 Sampling Distributions

23.3 Sampling Distributions COMMON CORE Locker LESSON Commo Core Math Stadards The studet is expected to: COMMON CORE S-IC.B.4 Use data from a sample survey to estimate a populatio mea or proportio; develop a margi of error through

More information

Measures of Spread: Standard Deviation

Measures of Spread: Standard Deviation Measures of Spread: Stadard Deviatio So far i our study of umerical measures used to describe data sets, we have focused o the mea ad the media. These measures of ceter tell us the most typical value of

More information

Should We Care How Long to Publish? Investigating the Correlation between Publishing Delay and Journal Impact Factor 1

Should We Care How Long to Publish? Investigating the Correlation between Publishing Delay and Journal Impact Factor 1 Should We Care How Log to Publish? Ivestigatig the Correlatio betwee Publishig Delay ad Joural Impact Factor 1 Jie Xu 1, Jiayu Wag 1, Yuaxiag Zeg 2 1 School of Iformatio Maagemet, Wuha Uiversity, Hubei,

More information

Statistics Lecture 13 Sampling Distributions (Chapter 18) fe1. Definitions again

Statistics Lecture 13 Sampling Distributions (Chapter 18) fe1. Definitions again fe1. Defiitios agai Review the defiitios of POPULATIO, SAMPLE, PARAMETER ad STATISTIC. STATISTICAL IFERECE: a situatio where the populatio parameters are ukow, ad we draw coclusios from sample outcomes

More information

Estimation and Confidence Intervals

Estimation and Confidence Intervals Estimatio ad Cofidece Itervals Chapter 9 McGraw-Hill/Irwi Copyright 2010 by The McGraw-Hill Compaies, Ic. All rights reserved. GOALS 1. Defie a poit estimate. 2. Defie level of cofidece. 3. Costruct a

More information

Chapter 21. Recall from previous chapters: Statistical Thinking. Chapter What Is a Confidence Interval? Review: empirical rule

Chapter 21. Recall from previous chapters: Statistical Thinking. Chapter What Is a Confidence Interval? Review: empirical rule Chapter 21 What Is a Cofidece Iterval? Chapter 21 1 Review: empirical rule Chapter 21 5 Recall from previous chapters: Parameter fixed, ukow umber that describes the populatio Statistic kow value calculated

More information

Statistical Analysis and Graphing

Statistical Analysis and Graphing BIOL 202 LAB 4 Statistical Aalysis ad Graphig Aalyzig data objectively to determie if sets of data differ ad the to preset data to a audiece succictly ad clearly is a major focus of sciece. We eed a way

More information

A longitudinal study of self-assessment accuracy

A longitudinal study of self-assessment accuracy The teachig eviromet A logitudial study of self-assessmet accuracy James T Fitzgerald, Casey B White & Larry D Gruppe Aim Although studies have examied medical studets ability to self-assess their performace,

More information

Sampling Distributions and Confidence Intervals

Sampling Distributions and Confidence Intervals 1 6 Samplig Distributios ad Cofidece Itervals Iferetial statistics to make coclusios about a large set of data called the populatio, based o a subset of the data, called the sample. 6.1 Samplig Distributios

More information

Lecture Outline. BIOST 514/517 Biostatistics I / Applied Biostatistics I. Paradigm of Statistics. Inferential Statistic.

Lecture Outline. BIOST 514/517 Biostatistics I / Applied Biostatistics I. Paradigm of Statistics. Inferential Statistic. BIOST 514/517 Biostatistics I / Applied Biostatistics I Kathlee Kerr, Ph.D. Associate Professor of Biostatistics iversity of Washigto Lecture 11: Properties of Estimates; Cofidece Itervals; Stadard Errors;

More information

Reporting Checklist for Nature Neuroscience

Reporting Checklist for Nature Neuroscience Correspodig Author: Mauscript Number: Mauscript Type: Galea NNA48318C Article Reportig Checklist for Nature Neurosciece # Figures: 4 # Supplemetary Figures: 2 # Supplemetary Tables: 1 # Supplemetary Videos:

More information

Bayesian Sequential Estimation of Proportion of Orthopedic Surgery of Type 2 Diabetic Patients Among Different Age Groups A Case Study of Government

Bayesian Sequential Estimation of Proportion of Orthopedic Surgery of Type 2 Diabetic Patients Among Different Age Groups A Case Study of Government Bayesia Sequetial Estimatio of Proportio of Orthopedic Surgery of Type Diabetic Patiets Amog Differet Age Groups A Case Study of Govermet Medical College, Jammu-Idia Roohi Gupta, Priyaka Aad ad *Rahul

More information

How is the President Doing? Sampling Distribution for the Mean. Now we move toward inference. Bush Approval Ratings, Week of July 7, 2003

How is the President Doing? Sampling Distribution for the Mean. Now we move toward inference. Bush Approval Ratings, Week of July 7, 2003 Samplig Distributio for the Mea Dr Tom Ilveto FREC 408 90 80 70 60 50 How is the Presidet Doig? 2/1/2001 4/1/2001 Presidet Bush Approval Ratigs February 1, 2001 through October 6, 2003 6/1/2001 8/1/2001

More information

Review for Chapter 9

Review for Chapter 9 Review for Chapter 9 1. For which of the followig ca you use a ormal approximatio? a) = 100, p =.02 b) = 60, p =.4 c) = 20, p =.6 d) = 15, p = 2/3 e) = 10, p =.7 2. What is the probability of a sample

More information

Lecture 19: Analyzing transcriptome datasets. Spring 2018 May 3, 2018

Lecture 19: Analyzing transcriptome datasets. Spring 2018 May 3, 2018 Lecture 19: Aalyzig trascriptome datasets Sprig 2018 May 3, 2018 Measurig the Trascriptome trascriptome: the mrnas expressed by a geome at ay give time (Abbott, 1999) Icludes protei codig trascripts ad

More information

Technical Assistance Document Algebra I Standard of Learning A.9

Technical Assistance Document Algebra I Standard of Learning A.9 Techical Assistace Documet 2009 Algebra I Stadard of Learig A.9 Ackowledgemets The Virgiia Departmet of Educatio wishes to express sicere thaks to J. Patrick Liter, Doa Meeks, Dr. Marcia Perry, Amy Siepka,

More information

Appendix C: Concepts in Statistics

Appendix C: Concepts in Statistics Appedi C. Measures of Cetral Tedecy ad Dispersio A8 Appedi C: Cocepts i Statistics C. Measures of Cetral Tedecy ad Dispersio Mea, Media, ad Mode I may real-life situatios, it is helpful to describe data

More information

Development Report of Powerful Acoustic Computing Environment

Development Report of Powerful Acoustic Computing Environment Developmet Report of Powerful Acoustic Computig Eviromet Takayuki asumoto echaical CAE Divisio VPD Group Cyberet Systems Co. td. 006 ANSYS Ic. ANSYS Ic. Proprietary Ageda Part: The Curret Status of Numerical

More information

Plantar Pressure Difference: Decision Criteria of Motor Relearning Feedback Insole for Hemiplegic Patients

Plantar Pressure Difference: Decision Criteria of Motor Relearning Feedback Insole for Hemiplegic Patients 22 4th Iteratioal Coferece o Bioiformatics ad Biomedical Techology IPCBEE vol.29 (22) (22) IACSIT Press, Sigapore Platar Pressure Differece: Decisio Criteria of Motor Relearig Feedback Isole for Hemiplegic

More information

Chapter 8 Descriptive Statistics

Chapter 8 Descriptive Statistics 8.1 Uivariate aalysis ivolves a sigle variable, for examples, the weight of all the studets i your class. Comparig two thigs, like height ad weight, is bivariate aalysis. (Which we will look at later)

More information

Guidance on the use of the Title Consultant Psychologist

Guidance on the use of the Title Consultant Psychologist Guidace o the use of the Title Cosultat Psychologist If you have problems readig this documet ad would like it i a differet format, please cotact us with your specific requiremets. Tel: 0116 2254 9568;

More information

Practical Basics of Statistical Analysis

Practical Basics of Statistical Analysis Practical Basics of Statistical Aalysis David Keffer Dept. of Materials Sciece & Egieerig The Uiversity of Teessee Koxville, TN 37996-2100 dkeffer@utk.edu http://clausius.egr.utk.edu/ Goveror s School

More information

5/7/2014. Standard Error. The Sampling Distribution of the Sample Mean. Example: How Much Do Mean Sales Vary From Week to Week?

5/7/2014. Standard Error. The Sampling Distribution of the Sample Mean. Example: How Much Do Mean Sales Vary From Week to Week? Samplig Distributio Meas Lear. To aalyze how likely it is that sample results will be close to populatio values How probability provides the basis for makig statistical ifereces The Samplig Distributio

More information

Gestalt Psychology. Chapter 12

Gestalt Psychology. Chapter 12 Chapter 12 Gestalt Psychology Cogitive Psychology, Third Editio by Kathlee M. Galotti Copyright 2004 by Wadsworth Publishig, a divisio of Thomso Learig The Gestalt revolt Aroud 1912. Structuralism i waig,

More information

Performance Improvement in the Bivariate Models by using Modified Marginal Variance of Noisy Observations for Image-Denoising Applications

Performance Improvement in the Bivariate Models by using Modified Marginal Variance of Noisy Observations for Image-Denoising Applications PROCEEDING OF WORLD ACADEM OF CIENCE, ENGINEERING AND ECHNOLOG VOLUME 5 APRIL 005 IN 307-6884 Performace Improvemet i the Bivariate Models by usig Modified Margial Variace of Noisy Observatios for Image-Deoisig

More information

Children and adults with Attention-Deficit/Hyperactivity Disorder cannot move to the beat

Children and adults with Attention-Deficit/Hyperactivity Disorder cannot move to the beat 1 SUPPLEMENTARY INFORMATION Childre ad adults with Attetio-Deficit/Hyperactivity Disorder caot move to the beat Frédéric Puyjariet 1, Valeti Bégel 1,2, Régis Lopez 3,4, Delphie Dellacherie 5,6, & Simoe

More information

DISTRIBUTION AND PROPERTIES OF SPERMATOZOA IN DIFFERENT FRACTIONS OF SPLIT EJACULATES*

DISTRIBUTION AND PROPERTIES OF SPERMATOZOA IN DIFFERENT FRACTIONS OF SPLIT EJACULATES* FERTILITY AND STERILITY Copyright 1972 by The Williams & Wilkis Co. Vol. 23, No.4, April 1972 Prited i U.S.A. DISTRIBUTION AND PROPERTIES OF SPERMATOZOA IN DIFFERENT FRACTIONS OF SPLIT EJACULATES* R. ELIASSON,

More information

Measuring the Ability to Identify One s Own Emotions: The Development and Initial Psychometric Evaluation of a Maximum-Performance Test

Measuring the Ability to Identify One s Own Emotions: The Development and Initial Psychometric Evaluation of a Maximum-Performance Test Sesatios Test 1 Measurig the Ability to Idetify Oe s Ow Emotios: The Developmet ad Iitial Psychometric Evaluatio of a Maximum-Performace Test Kimberly A. Barchard ad Jeifer Skeem Uiversity of Nevada, Las

More information

Measuring Dispersion

Measuring Dispersion 05-Sirki-4731.qxd 6/9/005 6:40 PM Page 17 CHAPTER 5 Measurig Dispersio PROLOGUE Comparig two groups by a measure of cetral tedecy may ru the risk for each group of failig to reveal valuable iformatio.

More information

Autism Awareness Education. April 2018

Autism Awareness Education. April 2018 Autism Awareess Educatio April 2018 What is Autism Autism is a wide-spectrum metal disorder that is talked about every day i health circles, but few really kow all the facts about it. Research cotiues

More information

Methodology CHAPTER OUTLINE

Methodology CHAPTER OUTLINE Methodology 2 CHAPTER OUTLINE LEARNING OBJECTIVES INTRODUCTION SOME FUNDAMENTALS Research methods ad statistics Carryig out quality research The role of theory i psychology DESIGNING EXPERIMENTS IN PSYCHOLOGY

More information

04/11/2014 YES* YES YES. Attitudes = Evaluation. Attitudes = Unique Cognitive Construct. Attitudes Predict Behaviour

04/11/2014 YES* YES YES. Attitudes = Evaluation. Attitudes = Unique Cognitive Construct. Attitudes Predict Behaviour CLICK HERE : EXAMINING IMPLICIT AND EXPLICIT ATTITUDES TOWARD RAPE AND SEXUALLY AGGRESSIVE BEHAVIOUR IN MEN RECRUITED ONLINE Chatal A. Herma, Kevi L. Nues, & Natasha Loricz October 31 st, 2014 Carleto

More information

Basic Requirements. of meeting cow herd production and profitability goals for the beef cattle enterprise.

Basic Requirements. of meeting cow herd production and profitability goals for the beef cattle enterprise. Basic Requiremets It is imperative that cattle producers have a adequate uderstadig of the basic utriet requiremets of the cow herd to make iformed ad effective utritio-related decisios. by Matt Hersom,

More information

Concepts Module 7: Comparing Datasets and Comparing a Dataset with a Standard

Concepts Module 7: Comparing Datasets and Comparing a Dataset with a Standard Cocepts Module 7: Comparig Datasets ad Comparig a Dataset with a Stadard Idepedece of each data poit Test statistics Cetral Limit Theorem Stadard error of the mea Cofidece iterval for a mea Sigificace

More information

CHAPTER 8 ANSWERS. Copyright 2012 Pearson Education, Inc. Publishing as Addison-Wesley

CHAPTER 8 ANSWERS. Copyright 2012 Pearson Education, Inc. Publishing as Addison-Wesley CHAPTER 8 ANSWERS Sectio 8.1 Statistical Literacy ad Critical Thikig 1 The distributio of radomly selected digits from to 9 is uiform. The distributio of sample meas of 5 such digits is approximately ormal.

More information

Meningococcal B Prevention Tools for Your Practice

Meningococcal B Prevention Tools for Your Practice Meigococcal B Prevetio Tools for Your Practice NAPNAP MeB Facts for HCPs Fast Facts Although ucommo, MeB is potetially fatal. 1 MeB symptoms progress quickly; death ca occur i 24 hours or less. MeB accouts

More information

An Approach for Type Synthesis of Overconstrained 1T2R Parallel Mechanisms

An Approach for Type Synthesis of Overconstrained 1T2R Parallel Mechanisms A Approach for Type Sythesis of Overcostraied 1T2R Parallel Mechaisms C. Dog 1, H. Liu 1, Q. Liu 1, T. Su 1, T. Huag 1, 2 ad D. G. Chetwyd 2 1 Key Laboratory of Mechaism Theory ad Equipmet Desig of State

More information

STATISTICAL ANALYSIS & ASTHMATIC PATIENTS IN SULAIMANIYAH GOVERNORATE IN THE TUBER-CLOSES CENTER

STATISTICAL ANALYSIS & ASTHMATIC PATIENTS IN SULAIMANIYAH GOVERNORATE IN THE TUBER-CLOSES CENTER March 3. Vol., No. ISSN 37-3 IJRSS & K.A.J. All rights reserved STATISTICAL ANALYSIS & ASTHMATIC PATIENTS IN SULAIMANIYAH GOVERNORATE IN THE TUBER-CLOSES CENTER Dr. Mohammad M. Faqe Hussai (), Asst. Lecturer

More information

RADIESSE Dermal Filler for the Correction of Moderate to Severe Facial Wrinkles and Folds, Such As Nasolabial Folds

RADIESSE Dermal Filler for the Correction of Moderate to Severe Facial Wrinkles and Folds, Such As Nasolabial Folds A PATIENT S GUIDE RADIESSE Dermal Filler for the Correctio of Moderate to Severe Facial Wrikles ad Folds, Such As Nasolabial Folds Read all the iformatio before you are treated with Radiesse dermal filler.

More information

Copy of: Proc. IEEE 1998 Int. Conference on Microelectronic Test Structures, Vol.11, March 1998

Copy of: Proc. IEEE 1998 Int. Conference on Microelectronic Test Structures, Vol.11, March 1998 Copy of: Proc. IEEE 998 It. Coferece o Microelectroic Test Structures, Vol., March 998 Wafer Level efect esity istributio Usig Checkerboard Test Structures Christopher Hess, Larg H. Weilad Istitute of

More information

Ovarian Cancer Survival

Ovarian Cancer Survival Dairy Products, Calcium, Vitami D, Lactose ad Ovaria Cacer: Results from a Pooled Aalysis of Cohort Studies Stephaie Smith-Warer, PhD Departmets of Nutritio & Epidemiology Harvard School of Public Health

More information

EDEXCEL NATIONAL CERTIFICATE UNIT 28 FURTHER MATHEMATICS FOR TECHNICIANS OUTCOME 1- ALGEBRAIC TECHNIQUES TUTORIAL 3 - STATISTICAL TECHNIQUES

EDEXCEL NATIONAL CERTIFICATE UNIT 28 FURTHER MATHEMATICS FOR TECHNICIANS OUTCOME 1- ALGEBRAIC TECHNIQUES TUTORIAL 3 - STATISTICAL TECHNIQUES EDEXCEL NATIONAL CERTIFICATE UNIT 8 FURTHER MATHEMATICS FOR TECHNICIANS OUTCOME 1- ALGEBRAIC TECHNIQUES TUTORIAL 3 - STATISTICAL TECHNIQUES CONTENTS Be able to apply algebraic techiques Arithmetic progressio

More information

Finite Element Simulation of a Doubled Process of Tube Extrusion and Wall Thickness Reduction

Finite Element Simulation of a Doubled Process of Tube Extrusion and Wall Thickness Reduction World Joural of Mechaics, 13, 3, 5- http://dx.doi.org/1.3/wjm.13.35 Published lie August 13 (http://www.scirp.org/joural/wjm) Fiite Elemet Simulatio of a Doubled Process of Tube Extrusio ad Wall Thickess

More information

Standard deviation The formula for the best estimate of the population standard deviation from a sample is:

Standard deviation The formula for the best estimate of the population standard deviation from a sample is: Geder differeces Are there sigificat differeces betwee body measuremets take from male ad female childre? Do differeces emerge at particular ages? I this activity you will use athropometric data to carry

More information

The relationship between hypercholesterolemia as a risk factor for stroke and blood viscosity measured using Digital Microcapillary

The relationship between hypercholesterolemia as a risk factor for stroke and blood viscosity measured using Digital Microcapillary Joural of Physics: Coferece Series PAPER OPEN ACCESS The relatioship betwee hypercholesterolemia as a risk factor for stroke ad blood viscosity measured usig Digital Microcapillary To cite this article:

More information

CEREC Omnicam: scanning simplicity.

CEREC Omnicam: scanning simplicity. C A D / C A M S Y S T EM S I N S T RU M EN T S H YG I EN E S Y S T EM S T R E AT M EN T CEN T ER S I M AG I N G S Y S T EM S C A D / C A M came r as. M ade t o i s p i r e cerec Omicam ad cerec Bluecam.

More information

What are minimal important changes for asthma measures in a clinical trial?

What are minimal important changes for asthma measures in a clinical trial? Eur Respir J 1999; 14: 23±27 Prited i UK ± all rights reserved Copyright #ERS Jourals Ltd 1999 Europea Respiratory Joural ISSN 0903-1936 What are miimal importat chages for asthma measures i a cliical

More information

SEIZURE SIGNALS SEPARATION USING CONSTRAINED TOPOGRAPHIC BLIND SOURCE SEPARATION

SEIZURE SIGNALS SEPARATION USING CONSTRAINED TOPOGRAPHIC BLIND SOURCE SEPARATION SEIZURE SIGALS SEPARATIO USIG COSTRAIED TOPOGRAPHIC BLID SOURCE SEPARATIO Mi Jig ad Saeid Saei Cetre of Digital Sigal Processig, Cardiff Uiversity Cardiff, CF24 3AA, South Wales, UK Email: {jigm, saeis}@cf.ac.uk

More information

The Nutritional Density Ratio Dilemma: Developing a Scale for Nutritional Value Paul D. Q. Campbell

The Nutritional Density Ratio Dilemma: Developing a Scale for Nutritional Value Paul D. Q. Campbell The Nutritioal Desity Ratio Dilemma: Developig a Scale for Nutritioal Value Paul D. Q. Campbell Shared by Paul D. Q. Campbell The author(s) would appreciate your feedback o this article. Click the yellow

More information

Chem 135: First Midterm

Chem 135: First Midterm Chem 135: First Midterm September 30 th, 2013 Please provide all aswers i the spaces provided. You are ot allowed to use a calculator for this exam, but you may use (previously disassembled) molecular

More information

Comparison of speed and accuracy between manual and computer-aided measurements of dental arch and jaw arch lengths in study model casts

Comparison of speed and accuracy between manual and computer-aided measurements of dental arch and jaw arch lengths in study model casts Compariso of speed ad accuracy betwee maual ad computeraided measuremets (Diah Wibisoo, et.al.) Compariso of speed ad accuracy betwee maual ad computeraided measuremets of detal arch ad jaw arch legths

More information

A Supplement to Improved Likelihood Inferences for Weibull Regression Model by Yan Shen and Zhenlin Yang

A Supplement to Improved Likelihood Inferences for Weibull Regression Model by Yan Shen and Zhenlin Yang A Supplemet to Improved Likelihood Ifereces for Weibull Regressio Model by Ya She ad Zheli Yag More simulatio experimets were carried out to ivestigate the effect of differet cesorig percetages o the performace

More information

Quantitative Evaluation of Stress Corrosion Cracking Based on Features of Eddy Current Testing Signals

Quantitative Evaluation of Stress Corrosion Cracking Based on Features of Eddy Current Testing Signals E-Joural of Advaced Maiteace Vol.9-2 (2017) 78-83 Japa Society of Maiteology Quatitative Evaluatio of Stress Corrosio Crackig Based o Features of Eddy Curret Testig Sigals Li WANG 1,* ad Zhemao CHEN 2

More information

Modified Early Warning Score Effect in the ICU Patient Population

Modified Early Warning Score Effect in the ICU Patient Population Lehigh Valley Health Network LVHN Scholarly Works Patiet Care Services / Nursig Modified Early Warig Score Effect i the ICU Patiet Populatio Ae Rabert RN, DHA, CCRN, NE-BC Lehigh Valley Health Network,

More information

Sec 7.6 Inferences & Conclusions From Data Central Limit Theorem

Sec 7.6 Inferences & Conclusions From Data Central Limit Theorem Sec 7. Ifereces & Coclusios From Data Cetral Limit Theorem Name: The Cetral Limit Theorem offers us the opportuity to make substatial statistical predictios about the populatio based o the sample. To better

More information

Acoustic Scene Classification by Ensembling Gradient Boosting Machine and Convolutional Neural Networks

Acoustic Scene Classification by Ensembling Gradient Boosting Machine and Convolutional Neural Networks Acoustic Scee Classificatio by Esemblig Gradiet Boostig Machie ad Covolutioal Neural Networks DCASE 2017 Eduardo Foseca, Rog Gog, Dmitry Bogdaov, Olga Slizovskaia, Emilia Gomez ad Xavier Serra Outlie Itroductio

More information

PDSS: The decision support system of diabetic patient for Public Health

PDSS: The decision support system of diabetic patient for Public Health Proceedigs of the 3rd Iteratioal Coferece o Idustrial Applicatio Egieerig 5 PDSS: The decisio support system of diabetic patiet for Public Health Bejapuk Jogmuewai, Kailas Bumrugchat, Papo kaewhi Iformatics

More information

An Automatic Denoising Method with Estimation of Noise Level and Detection of Noise Variability in Continuous Glucose Monitoring

An Automatic Denoising Method with Estimation of Noise Level and Detection of Noise Variability in Continuous Glucose Monitoring Preprit, 11th IFAC Symposium o Dyamics ad Cotrol of Process Systems, icludig Biosystems Jue 6-8, 16. NTNU, Trodheim, Norway A Automatic Deoisig Method with Estimatio of Noise Level ad Detectio of Noise

More information

S3: Ultrasensitization is Preserved for Transient Stimuli

S3: Ultrasensitization is Preserved for Transient Stimuli S3: Ultrasesitizatio is Preserved for Trasiet Stimuli I the followig we show that ultrasesitizatio is preserved (albeit weaeed) upo trasiet stimulatio (e.g. due to receptor dowregulatio) as log as the

More information

Open Research Online The Open University s repository of research publications and other research outputs

Open Research Online The Open University s repository of research publications and other research outputs Ope Research Olie The Ope Uiversity s repository of research publicatios ad other research outputs Exploitig coceptual spaces for otology itegratio Coferece or Workshop Item How to cite: Dietze, Stefa

More information

1 Barnes D and Lombardo C (2006) A Profile of Older People s Mental Health Services: Report of Service Mapping 2006, Durham University.

1 Barnes D and Lombardo C (2006) A Profile of Older People s Mental Health Services: Report of Service Mapping 2006, Durham University. The Natioal Audit Office udertook a self-assessmet cesus of Commuity Metal Health Teams for Older People (CMHTs) betwee September ad December 2006. The overall fidigs are preseted i the Natioal Audit Office

More information

AND FEEDBACK SEEKING TOWARDS ENTREPRENEURIAL PERFORMANCE OF STUDENTS AT CIPUTRA UNIVERSITY

AND FEEDBACK SEEKING TOWARDS ENTREPRENEURIAL PERFORMANCE OF STUDENTS AT CIPUTRA UNIVERSITY AND EEDBACK EEKIN TOWARD ENTREPRENEURIAL PERORMANCE O TUDENT AT CIPUTRA UNIVERITY Wia Christia, Herry Purwoko Lecturer at Ciputra Uiversity, urabaya, Idoesia Lecturer at Ciputra Uiversity, urabaya, Idoesia

More information

The US population aged 75 years or more has

The US population aged 75 years or more has 551 Blood Pressure Chage ad Survival After Age 75 Robert D. Lager, Michael H. Criqui, Elizabeth L. Barrett-Coor, Melville R. Klauber, Theodore G. Gaiats Higher diastolic pressure predicted better survival

More information

Primary: To assess the change on the subject s quality of life between diagnosis and the first 3 months of treatment.

Primary: To assess the change on the subject s quality of life between diagnosis and the first 3 months of treatment. Study No.: AVO112760 Title: A Observatioal Study To Assess The Burde Of Illess I Prostate Cacer Patiets With Low To Moderate Risk Of Progressio Ratioale: Little data are available o the burde of illess

More information

Statistics for Managers Using Microsoft Excel Chapter 7 Confidence Interval Estimation

Statistics for Managers Using Microsoft Excel Chapter 7 Confidence Interval Estimation Statistics for Maagers Usig Microsoft Excel Chapter 7 Cofidece Iterval Estimatio 1999 Pretice-Hall, Ic. Chap. 7-1 Chapter Topics Cofidece Iterval Estimatio for the Mea (s Kow) Cofidece Iterval Estimatio

More information

Objectives. Types of Statistical Inference. Statistical Inference. Chapter 19 Confidence intervals: Estimating with confidence

Objectives. Types of Statistical Inference. Statistical Inference. Chapter 19 Confidence intervals: Estimating with confidence Types of Statistical Iferece Chapter 19 Cofidece itervals: The basics Cofidece itervals for estiatig the value of a populatio paraeter Tests of sigificace assesses the evidece for a clai about a populatio.

More information

Relationship between Established Breast Cancer Risk Factors and Risk of Seven Different Histologic Types of Invasive Breast Cancer

Relationship between Established Breast Cancer Risk Factors and Risk of Seven Different Histologic Types of Invasive Breast Cancer 946 Relatioship betwee Established Breast Cacer Risk Factors ad Risk of Seve Differet Histologic Types of Ivasive Breast Cacer Christopher I. Li, 1 Jaet R. Dalig, 1 Kathlee E. Maloe, 1 Leslie Berstei,

More information

Teacher Manual Module 3: Let s eat healthy

Teacher Manual Module 3: Let s eat healthy Teacher Maual Module 3: Let s eat healthy Teacher Name: Welcome to FLASH (Fu Learig Activities for Studet Health) Module 3. I the Uited States, more studets are developig type 2 diabetes tha ever before.

More information

Introduction. The Journal of Nutrition Methodology and Mathematical Modeling

Introduction. The Journal of Nutrition Methodology and Mathematical Modeling The Joural of Nutritio Methodology ad Mathematical Modelig The Populatio Distributio of Ratios of Usual Itakes of Dietary Compoets That Are Cosumed Every Day Ca Be Estimated from Repeated 24-Hour Recalls

More information

PACIFICA M.A. IN COUNSELING PSYCHOLOGY. With Emphasis in Marriage and Family Therapy, Professional Clinical Counseling, and Depth Psychology

PACIFICA M.A. IN COUNSELING PSYCHOLOGY. With Emphasis in Marriage and Family Therapy, Professional Clinical Counseling, and Depth Psychology PACIFICA g r a d u a t e i s t i t u t e M.A. IN COUNSELING PSYCHOLOGY With Emphasis i Marriage ad Family Therapy, Professioal Cliical Couselig, ad Depth Psychology PACIFICA GRADUATE INSTITUTE 249 LAMBERT

More information

A Simulated Global Neuronal Workspace with Stochastic Wiring

A Simulated Global Neuronal Workspace with Stochastic Wiring A Simulated Global Neuroal Workspace with Stochastic Wirig Dusti Coor ad Murray Shaaha Departmet of Computig, Imperial College Lodo, 180 Quee s Gate, Lodo SW7 2AZ, Eglad, UK dusti.coor02@imperial.ac.uk;

More information

Clinical Usefulness of Very High and Very Low Levels of C-Reactive Protein Across the Full Range of Framingham Risk Scores

Clinical Usefulness of Very High and Very Low Levels of C-Reactive Protein Across the Full Range of Framingham Risk Scores Cliical Usefuless of Very High ad Very Low Levels of C-Reactive Protei Across the Full Rage of Framigham Risk Scores Paul M Ridker, MD, MPH; Nacy Cook, ScD Backgroud High-sesitivity C-reactive protei (hscrp)

More information

Outline. Neutron Interactions and Dosimetry. Introduction. Tissue composition. Neutron kinetic energy. Neutron kinetic energy.

Outline. Neutron Interactions and Dosimetry. Introduction. Tissue composition. Neutron kinetic energy. Neutron kinetic energy. Outlie Neutro Iteractios ad Dosimetry Chapter 16 F.A. Attix, Itroductio to Radiological Physics ad Radiatio Dosimetry Neutro dosimetry Thermal eutros Itermediate-eergy eutros Fast eutros Sources of eutros

More information

EFSA Guidance for BMD analysis Fitting Models & Goodness of Fit

EFSA Guidance for BMD analysis Fitting Models & Goodness of Fit EFSA Guidace for BMD aalysis Fittig Models & Goodess of Fit 1 st March 2017 OUTLINE Geeral priciples of model fittig & goodess of fit Cotiuous dose-respose data Cell proliferatio (CP) data Cadidate models

More information

Minimum skills required by children to complete healthrelated quality of life instruments for asthma: comparison of measurement properties

Minimum skills required by children to complete healthrelated quality of life instruments for asthma: comparison of measurement properties Eur Respir J 1997; 10: 225 24 DOI: 10.113/09031936.97.1010225 Prited i UK - all rights reserved Copyright ERS Jourals Ltd 1997 Europea Respiratory Joural ISSN 0903-1936 Miimum skills required by childre

More information

Chapter 8 Student Lecture Notes 8-1

Chapter 8 Student Lecture Notes 8-1 Chapter 8 tudet Lecture Notes 8-1 Basic Busiess tatistics (9 th Editio) Chapter 8 Cofidece Iterval Estimatio 004 Pretice-Hall, Ic. Chap 8-1 Chapter Topics Estimatio Process Poit Estimates Iterval Estimates

More information

Chapter - 8 BLOOD PRESSURE CONTROL AND DYSLIPIDAEMIA IN PATIENTS ON DIALYSIS

Chapter - 8 BLOOD PRESSURE CONTROL AND DYSLIPIDAEMIA IN PATIENTS ON DIALYSIS Chapter - BLOOD PRESSURE CONTROL AND DYSLIPIDAEMIA IN PATIENTS ON DIALYSIS S. Prasad Meo Hooi Lai Seog Lee Wa Ti Suita Bavaada ST REPORT OF THE MALAYSIAN DIALYSIS AND TRANSPLANT REGISTRY SECTION.: BLOOD

More information

Intro to Scientific Analysis (BIO 100) THE t-test. Plant Height (m)

Intro to Scientific Analysis (BIO 100) THE t-test. Plant Height (m) THE t-test Let Start With a Example Whe coductig experimet, we would like to kow whether a experimetal treatmet had a effect o ome variable. A a imple but itructive example, uppoe we wat to kow whether

More information

Rheological Characterization of Fiber Suspensions Prepared from Vegetable Pulp and Dried Fibers. A Comparative Study.

Rheological Characterization of Fiber Suspensions Prepared from Vegetable Pulp and Dried Fibers. A Comparative Study. ANNUAL TRANSACTIONS OF THE NORDIC RHEOLOGY SOCIETY, VOL. 3, 5 Rheological Characterizatio of Fiber Suspesios Prepared from Vegetable Pulp ad Dried Fibers. A Comparative Study. Elea Bayod, Ulf Bolmstedt

More information

Open-Source Programming of Cardiovascular Pressure-Flow Dynamics Using SimPower Toolbox in Matlab and Simulink

Open-Source Programming of Cardiovascular Pressure-Flow Dynamics Using SimPower Toolbox in Matlab and Simulink The Ope Pacig, Electrophysiology & Therapy Joural, 1, 3, 55-59 55 Ope Access Ope-Source Programmig of Cardiovascular Pressure-Flow Dyamics Usig SimPower Toolbox i Matlab ad Simulik Ofer Barea * Departmet

More information

Pilot and Exploratory Project Support Grant

Pilot and Exploratory Project Support Grant KEY DATES LETTERS OF INTENT DUE November 3, 2014 5:00 pm est FULL PROPOSAL INVITATIONS November 17, 2014 FULL PROPOSAL DEADLINE Jauary 15, 2015 5:00 pm est NOTIFICATION OF AWARDS April, 2015 Pilot ad Exploratory

More information

CMP: Data Mining and Statistics Within the Health Services 19/02/2010

CMP: Data Mining and Statistics Within the Health Services 19/02/2010 CMP: Data Miig ad Statistics Withi the Health Services 19//1 Data Miig ad Statistics Withi the Health Services Cotet Data Miig Esemble for Predictig Osteoporosis Dr. Wejia Wag School of Computig Scieces

More information

Epilepsy and Family Dynamics

Epilepsy and Family Dynamics Epilepsy ad Family Dyamics BC Epilepsy Society November 15, 2010 Guests: Susa Murphy, Registered Nurse, Paret Rita Marchildo, Child Life Specialist, Paret Speakers: Audrey Ho PhD., R.Psych Josef Zaide

More information

Neural processing of biomedical data for improving driving safety

Neural processing of biomedical data for improving driving safety Modellig i Medicie ad Biology VI 213 Neural processig of biomedical data for improvig drivig safety F. Barcelloa 1, F. Filippi 1, M. Paella 2, A. M. Bersai 3 & A. Alessadrii 1. 1 D.I.T.S. Departmet, Uiversity

More information

Event detection. Biosignal processing, S Autumn 2017

Event detection. Biosignal processing, S Autumn 2017 Evet detectio Biosigal processig, 573S Autum 07 ECG evet detectio P wave: depolarizatio of the atrium QRS-complex: depolarizatio of vetricle T wave: repolarizatio of vetricle Each evet represets oe phase

More information

Improving the Bioanalysis of Endogenous Bile Acids as Biomarkers for Hepatobiliary Toxicity using Q Exactive Benchtop Orbitrap?

Improving the Bioanalysis of Endogenous Bile Acids as Biomarkers for Hepatobiliary Toxicity using Q Exactive Benchtop Orbitrap? Troy Voelker, Mi Meg Tadem Labs, Salt Lake City, UT Kevi Cook, Patrick Beett Thermo Fisher Scietific, Sa Jose, CA Improvig the Bioaalysis of Edogeous Bile Acids as Biomarkers for Hepatobiliary Toxicity

More information

DEGRADATION OF PROTECTIVE GLOVE MATERIALS EXPOSED TO COMMERCIAL PRODUCTS: A COMPARATIVE STUDY OF TENSILE STRENGTH AND GRAVIMETRIC ANALYSES

DEGRADATION OF PROTECTIVE GLOVE MATERIALS EXPOSED TO COMMERCIAL PRODUCTS: A COMPARATIVE STUDY OF TENSILE STRENGTH AND GRAVIMETRIC ANALYSES Califoria State Uiversity, Sa Berardio CSUSB ScholarWorks Electroic Theses, Projects, ad Dissertatios Office of Graduate Studies 9-2014 DEGRADATION OF PROTECTIVE GLOVE MATERIALS EXPOSED TO COMMERCIAL PRODUCTS:

More information

Visual Acuity Screening of Children 6 Months to 3 Years of Age

Visual Acuity Screening of Children 6 Months to 3 Years of Age Visual Acuity Screeig of Childre Moths to Years of Age Velma Dobso,* Deborah Salem,t D. Luisa Mayer, Cythia Moss, ad S. Lawso Sebris The operat preferetial lookig () procedure allows a behavioral estimate

More information

The Northern Trust Experience

The Northern Trust Experience The Norther Trust Experiece ACCESS. EXPERTISE. SERVICE. Keepig It Together: Helpig the Family Stay Effective i Ucertai Times November 5, 2010 A Freel Director, Family Educatio ad Goverace Services 2010

More information

The Efficiency of the Denver Developmental Screening Test with Rural Disadvantaged Preschool Children 1

The Efficiency of the Denver Developmental Screening Test with Rural Disadvantaged Preschool Children 1 Joural of Pediatric Psychology, Vol. 8, No. 3, 1983 The Efficiecy of the Dever Developmetal Screeig Test with Rural Disadvataged Preschool Childre 1 Deis C. Harper 2 ad David P. Wacker Departmet of Pediatrics,

More information

Laxmi Shaw Department of Applied Electronics and Instrumentation Silicon Institute of Technology, Bhubaneswar, Orissa, India

Laxmi Shaw Department of Applied Electronics and Instrumentation Silicon Institute of Technology, Bhubaneswar, Orissa, India HMM BASED PARKINSON S DETECTION BY ANALYSING SYMBOLIC POSTURAL GAIT IMAGE SEQUENCES Laxmi Shaw Departmet of Applied Electroics ad Istrumetatio Silico Istitute of Techology, Bhubaeswar, Orissa, Idia laxmishaw1983@gmail.com

More information

COMBINATORIAL ON/OFF MODEL FOR OLFACTORY CODING. A.Koulakov 1, A.Gelperin 2, and D.Rinberg 2

COMBINATORIAL ON/OFF MODEL FOR OLFACTORY CODING. A.Koulakov 1, A.Gelperin 2, and D.Rinberg 2 COMBINATORIAL ON/OFF MODEL FOR OLFACTORY CODING A.Koulakov, A.Gelperi 2, ad D.Riberg 2. Cold Sprig Harbor Laboratory, Cold Sprig Harbor, NY, 724 USA 2. Moell Chemical Seses Ctr., Philadelphia, PA, 904

More information

An algorithm for prioritizing the maintenance of power transformers

An algorithm for prioritizing the maintenance of power transformers Eergy Productio ad Maagemet i the 21st Cetury, Vol. 1 335 A algorithm for prioritizig the maiteace of power trasformers I. V. Davideko & E. D. Halikova Ural Power Egieerig of Ural Federal Uiversity, Russia

More information

GSK Medicine: Study Number: Title: Rationale: Study Period: Objectives: Indication: Study Investigators/Centers: Research Methods:

GSK Medicine: Study Number: Title: Rationale: Study Period: Objectives: Indication: Study Investigators/Centers: Research Methods: The study listed may iclude approved ad o-approved uses, mulatios or treatmet regimes. The results reported i ay sigle study may ot reflect the overall results obtaied o studies of a product. Bee prescribig

More information

Chapter 1 Introduction

Chapter 1 Introduction Chapter 1 Itroductio 1 The olefi metathesis reactio is a powerful sythetic tool that scrambles the carbo atoms of carbo carbo double bods ad creates ew carbo carbo double bods. The mechaism of the reactio

More information

Simple intervention to improve detection of hepatitis B and hepatitis C in general practice

Simple intervention to improve detection of hepatitis B and hepatitis C in general practice Simple itervetio to improve detectio of hepatitis B ad hepatitis C i geeral practice Zayab al-lami (GP-Birmigham) Co-authors:-Sarah Powell, Sally Bradshaw, Amada Lambert, David Mutimer ad Adrew Rouse Presetatio

More information

Rana M Zeina* PhD, 2, Laila AL- Ayadhi, MBBS, PhD 3, Shahid basher PhD.

Rana M Zeina* PhD, 2, Laila AL- Ayadhi, MBBS, PhD 3, Shahid basher PhD. Research Ivety: Iteratioal Joural Of Egieerig Ad Sciece Vol.4, Issue 11 (November2014), PP 35-39 Iss (e): 2278-4721, Iss (p):2319-6483, www.researchivety.com Brief report o usig europsychological computerized

More information