Hello,
I have an experimental design where I would like to use separate ranking
events to predict an independent ranking event. I have been using function
clmm in the ordinal library but now realize that I am violating one of the
assumptions. I have included a subset of the data.
I am looking at how undergraduate students perceive the architecture and
biological correctness of box and arrow models when deciding on a model's
overall quality. I constructed 9 models with 3 levels of complexity and 3
levels of correctness. Students performed 3 tasks: arrange the models
according to quality, correctness, and complexity. In the included data, a
placement of 9 represents the best, most correct, or most complex. For
example, Pink is often placed as the worst model.
I want to determine whether students are more likely to choose quality
models based on correctness or complexity. I have calculated Kendall’s tau
statistic, but am interested in a regression analysis where Overall Quality
is the response variable and Correctness and Complexity are predictor
variables. My current approach uses ordinal logistic regression with mixed
effects. However, once a student ranks a single model, there is not an
equal probability of each model being placed at any other rank. For
example, if student_026 thinks Pink is the least correct and ranks it 1,
then other models can only be placed in positions 2-9 (not all 9). This
continues for each ranking. Ordinal regression assumes you have the same
set of choices at every step.
Has anyone encountered this situation when using ranked data as predictors
and response? Is there an R package or function that handles this type of
analysis?
Thank you.
Student Color Quality Correctness Complexity
Student_026 Pink 1 1 6
Student_026 Brown 3 2 5
Student_026 Green 2 3 4
Student_026 Black 8 7 2
Student_026 Yellow 9 8 1
Student_026 Aqua 7 9 3
Student_026 Purple 4 4 9
Student_026 Orange 6 5 7
Student_026 Red 5 6 8
Student_036 Pink 1 1 1
Student_036 Brown 2 2 3
Student_036 Green 3 7 2
Student_036 Black 7 3 6
Student_036 Yellow 8 6 4
Student_036 Aqua 9 8 5
Student_036 Purple 4 4 8
Student_036 Orange 5 5 7
Student_036 Red 6 9 9
--
Joe Dauer
Research Associate
Dept. of Plant Biology
Michigan State University
jdauer@msu.edu
[[alternative HTML version deleted]]
