It appears that the outfit t statistics for tam.fit on a tam.jml object are too small by approximately a factor of 2.
The attached figure shows the TAM computed values of infit/outfit (on the x axis) and their TAM computed t transformations (on the y axis) for MMLE and JMLE models run on the same dataset. Note that the red dots (JMLE outfit) appear to be from a very different distribution than the other colors (MMLE infit/outfit and JMLE infit). Per my understanding, the t transformation should be roughly identical for all of these infit/outfit statistics because they are trained on the same model (same df, similar variance).
All values come from running tam.fit on, respectively, an object of either tam.mml or tam.jml.
I am happy to try to formulate a MWE if that will be helpful. Please feel free to reach out at josephine.meyer@colorado.edu. I am currently running R version 4.2.3 on macOS Catalina, TAM version 4.1-4.
Thanks!
It appears that the outfit t statistics for tam.fit on a tam.jml object are too small by approximately a factor of 2.
The attached figure shows the TAM computed values of infit/outfit (on the x axis) and their TAM computed t transformations (on the y axis) for MMLE and JMLE models run on the same dataset. Note that the red dots (JMLE outfit) appear to be from a very different distribution than the other colors (MMLE infit/outfit and JMLE infit). Per my understanding, the t transformation should be roughly identical for all of these infit/outfit statistics because they are trained on the same model (same df, similar variance).
All values come from running tam.fit on, respectively, an object of either tam.mml or tam.jml.
I am happy to try to formulate a MWE if that will be helpful. Please feel free to reach out at josephine.meyer@colorado.edu. I am currently running R version 4.2.3 on macOS Catalina, TAM version 4.1-4.
Thanks!
