I'm developing a short version of a validated questionnaire, using Rumm 2030, and have a final model including 13 items. I wanted to assess if any response dependence is found by looking into the residual correlation between the items.
As I understand it: "Any residual correlation value of >0.2 above the average of all item residual correlations reflects response dependence" (Christensen KB, Makransky G, Horton M. Critical Values for Yen's Q(3): Identification of Local Dependence in the Rasch Model Using Residual Correlations. Appl Psychol Meas. 2017;41(3):178-194.). Now to my questions:
1) when calculating the mean of the correlation matrix, should I keep all the 1 in the diagonal of the matrix? (The mean is 0.062 if I keep all the 1, and -0.07 if I remove all the 1 in the diagonal)
2) should I take into account whether the correlations are positive or negative? (if I get a cut-off value of 0.13, does this mean that a correlation of -0.15 is ok or not?)
Thank you.
Hi, I hope this is still useful. I calculated the average of the residual correlation example Christensen et al., (2017) present in Table 1. In the article, they say that the average of that residual correlation matrix is -.12. I calculated the average with negative values and without the diagonal and gat the same -.12 value.
Best,
JP
- Badges
- Science topic
- Similar topics
- Human-Computer Interaction
- Collaboration