I am no expert in the field but I've been trying to fit T-IRT models to a rank-type personality test.
I built blocks from my data frame, and then used make_TIRT_data to build the dataset. Originally lower values represent higher preference (within a triplet) and hence I chose an option of direction = smaller for fits done with fit_TIRT_lavaan and the same object was used in fit_TIRT_mplus. To my surprise predicted values as negatively correlated between the two solutions (with correlations of .99).
When I modified the TIRT_data option to direction = larger and re-run the estimation in mplus, the correlations are positive.
From the above I assumed that one of the function fit_TIRT_lavaan or fit_TIRT_mplus might have some error in it.
But I can't tell which one.
I am no expert in the field but I've been trying to fit T-IRT models to a rank-type personality test.
I built blocks from my data frame, and then used
make_TIRT_datato build the dataset. Originally lower values represent higher preference (within a triplet) and hence I chose an option ofdirection = smallerfor fits done withfit_TIRT_lavaanand the same object was used infit_TIRT_mplus. To my surprise predicted values as negatively correlated between the two solutions (with correlations of .99). When I modified the TIRT_data option todirection = largerand re-run the estimation in mplus, the correlations are positive.From the above I assumed that one of the function
fit_TIRT_lavaanorfit_TIRT_mplusmight have some error in it. But I can't tell which one.