Improving Measurement of Ambiguity Tolerance (AT) Among Teacher Candidates Kent Rittschof Department of Curriculum, Foundations, & Reading
What is Ambiguity Tolerance (AT) and why should it be measured? This psychological construct is sometimes referred to as Tolerance for Ambiguity or TA (Frenkel-Brunswik, 1949) Budner (1962) defined AT as the tendency to perceive ambiguous situations as desirable and intolerance for ambiguity (AKA uncertainty avoidance ) as the tendency to perceive (i.e. interpret) ambiguous situations as sources of threat
What is Ambiguity Tolerance (AT) and why should it be measured? AT can be informative, interesting, and potentially valuable within investigations of teaching and learning. Measurement of the construct has been evolving and improving since 1949 Preference to examine AT s current measurement status (with educators) before investing in further research.
Ambiguity Tolerance (AT) Research Findings Examples Teachers who had a higher AT tended to use higher cognitive levels of verbal responses in their teaching (Peters and Amburgey, 1982). AT investigations supported learning that involves complex problems, novel transfer using new examples, divergent learning tasks, and brainstorming (Jonassen & Grabowski, 1993). AT among teachers has correlated (r =.59) with a constructivist teaching orientation (Gottleib, 2006). Negative AT Associations: Dogamatism, Authoritarianism
Measurement of AT MSTAT-I (McLain, 1993) 22 item, Likert MSTAT-II (McLain, 2009) 13 item, Likert Stimulus (Item) Types oambiguous Stimuli General ouncertain Stimuli onovel Stimuli oinsoluable / Illogical Stimuli oexample Item I try to avoid situations that are ambiguous
MSTAT-I, MSTAT-II (Yellow) MSTAT-I (McLain, 1993) Ambiguity Tolerance Scale 1. I don t tolerate ambiguous situations very well. 1 2. I find it difficult to respond when faced with an unexpected event. 3. I don t think new situations are any more threatening than familiar situations. 4. I m drawn to situations that can be interpreted in more than one way. 5. I would rather avoid solving a problem that must be viewed from several different perspectives. 2 6. I try to avoid situations that are ambiguous. 3 7. I am good at managing unpredictable situations. 8. I prefer familiar situations to new ones. 4 9. Problems that cannot be considered from just one point of view are a little threatening. 5 10. I avoid situations that are too complicated for me to easily understand. 6 11. I am tolerant of ambiguous situations. 7 12. I enjoy tackling problems that are complex enough to be ambiguous. 8 13. I try to avoid problems that don t seem to have only one best solution. 9 14. I often find myself looking for something new, rather than trying to hold things constant in my life. 15. I generally prefer novelty over familiarity. 10 16. I dislike ambiguous situations. 11 17. Some problems are so complex that just trying to understand them is fun. 18. I have little trouble coping with unexpected events. 19. I pursue problem situations that are so complex some people call them mind boggling. 20. I find it hard to make a choice when the outcome is uncertain. 12 21. I enjoy an occasional surprise. 22. I prefer a situation in which there is some ambiguity. 13
Rasch Rating-Scale Measurement Model Contemporary, confirmatory, measurement model from the Rasch family of models. Useful diagnostic and visualization tools. Allows focus on items and people together Effective with modest sample sizes compared with other latent trait / IRT models. Log Odds of response Person Measure Item Difficulty.5 Prob. Threshold ln P nik P ni(k 1) = B n D i F k
Rasch IRT Software Examples Winsteps (www.winsteps.com/) Facets (www.winsteps.com/) jmetric (www.itemanalysis.com/) RUMM2030 (www.rummlab.com.au) (www.rasch-analysis.com/) ConQuest (www.acer.edu.au/conquest) R (cran.rproject.org/web/views/psychometrics.html) Software Directory www.rasch.org/software.htm
Measurement Diagnostics Demonstrated Reliability (0 to1) Separation (SE Units) Sample Targeting (Variable Maps) Person Fit (Zstd, Pathway Plot) Item Fit (Zstd, Pathway Plot) Dimensionality (Percentage Variance) Item Polarity (-1 to 1) Category Functioning (Cat. Prob Curves)
6 Calibrations among 4 versions 1. MSTAT-I (22 items) 22 2. MSTAT-II a (13 items) 13a 3. MSTAT-II b (13 items) 13b 4. MSTAT-U a (9 items) 9a 5. MSTAT-U b (9 items) 9b 6. MSTAT-General (5 items) 5 Will compare across all 6 calibrations on Reliability, Separation, Sample Targeting, Person Fit, Item Fit, Dimensionality, Item Polarity, Category Functioning.
Person (Test) Reliability and Separation Calib. Reliability Separation 22.87 2.65 13a.83 2.22 13b.84 2.30 9a.85 2.42 9b.86 2.44 5.81 2.09
Sample Targeting Variable Map MSTAT-I 22
Sample Targeting 22 compared with 13
Variable Map 13 Item Targeting of Items and Persons
9 items with 7 rating categories versus 5
9 items 5 rating categories versus 5 items 5 rating categories
22 Item -- Person Outfit Zstd Person 3 2 Less Measures More 1 0-1 -2 Overfit t Outfit Zstd Underfit -6-4 -2 0 2 4 6 8 10
22 Item Item Outfit Zstd 1 Item 0 Less Measures More -1-2 -6-4 -2 0 2 4 6 8 Overfit t Outfit Zstd Underfit
13 Item Item Outfit Zstd 1 Item Less Measures More 0-1 -6-4 -2 0 2 4 6 8 10 Overfit t Outfit Zstd Underfit
13 Item Item Outfit Zstd 1 Item Less Measures More 0-1 -4-2 0 2 4 6 8 Overfit t Outfit Zstd Underfit
9 Item, 7 category Item Outfit Zstd 2 Item 1 Less Measures More 0-1 -4-2 0 2 4 Overfit t Outfit Zstd Underfit
9 Item, 5 category Item Outfit Zstd 2 Item 1 Less Measures More 0-1 -4-2 0 2 4 Overfit t Outfit Zstd Underfit
5 Item, 5 category Item Outfit Zstd 1 Item Less Measures More 0-1 -4-2 0 2 Overfit t Outfit Zstd Underfit
Dimensionality: Variance (Principal-Components of Residuals) Calib. Measures First Contrast 22 35.5% 9.3% 13a 37.1% 9.9% 13b 39.0% 9.6% 9a 48.1% 9.6% 9b 48.3% 8.9% 5 52.6% 15.6%
Polarity: Point Measure Correlation 13 Item
Category Structure Analysis 13 Item
Category Probability Curves, 13 Item (Using Category Level Symbols)
Category Probability Curves 13 Item
Category Probability Curves 5 Item
Some Findings and Interpretations Two groupings of people are indicated by reliability and separation levels across calibrations of instrument versions. (Good! Though 3 groupings would have been better) Reliability/Separation declined very little with reduction to 13 and 9 item instruments. (Very Good!)
Some Findings and Interpretations Sample targeting across all calibrations indicates a suitable instrument match to this group of educators (Very Good!) Sample targeting redundancy declined with 13 item version but distribution coverage did not decline much. (Very Good!) 9 item and 5 item versions resulted in reduced range in distribution coverage (not good, but the expected cost)
Some Findings and Interpretations 13 item, 9 item, and 5 item versions showed respectively improved fit and uni-dimensionality (Very Good!) Dimensionality analysis of 13 and 9 item versions indicated a dominant dimension but also secondary dimensions (OK, but not Ideal) 5 category rather than 7 category Likert scale was supported by Category Function analysis (Try 5 category next time)
Some Findings and Interpretations Teacher candidates were normally distributed on Ambiguity Tolerance (AT) measures, with a range of 3.34 logits and a standard deviation of.60 logits, compared with MSTAT-II items range of 1.08 logits and a standard deviation of.32 logits on endorsement difficulty (Interesting. Useful for comparisons with other samples in future analyses)
Thank You. Questions? kent_r@georgiasouthern.edu