Strange discrepancy between multipleGroup function and IRTPRO

40 views
Skip to first unread message

Nikos Tsigilis

unread,
May 7, 2017, 10:31:33 AM5/7/17
to mirt-package
I am fiting a 2PL model for binary data to two groups using mirt multipleGroup function and IRTPRO.  Although I get EXACTLY the same fit indices and the same discrimination and difficulty parameters for males I get different results for the females. What might be the problem?
Below you can find the selected results from mirt and IRTPRO.  I have also included the first three items' parameters for each group. 

Log-likelihood = -11054.14
Estimated parameters: 42 
AIC = 22192.28; AICc = 22194.22
BIC = 22425.45; SABIC = 22292.02
G2 (981) = 1884.29, p = 0
RMSEA = 0.022, CFI = 0.687, TLI = 0.679

$Males
$items
         a1      d g u
y2co  1.337 -1.464 0 1
y5co  0.854 -0.266 0 1
y8co  1.117 -0.225 0 1

$Females
$items
         a1      d g u
y2co  1.501 -1.958 0 1
y5co  1.015 -0.593 0 1
y8co  0.876 -0.536 0 1

2PL Model Item Parameter Estimates for Group 1, logit: + c or a(θb)   (Back to TOC)
Item Label a s.e. c s.e. b s.e.
1 y2co 2 1.34 0.15 1 -1.46 0.11 1.09 0.10
2 y5co 4 0.85 0.10 3 -0.27 0.07 0.31 0.09
3 y8co 6 1.12 0.12 5 -0.23 0.08 0.20 0.07

2PL Model Item Parameter Estimates for Group 2, logit: + c or a(θb)   (Back to TOC)
Item Label a s.e. c s.e. b s.e.
1 y2co 22 1.67 0.39 21 -1.33 0.48 0.80 0.24
2 y5co 24 1.13 0.32 23 -0.17 0.33 0.15 0.29
3 y8co 26 0.98 0.28 25 -0.17 0.29 0.18 0.28

Likelihood-based Values and Goodness of Fit Statistics   (Back to TOC)
Statistics based on the loglikelihood
-2loglikelihood: 22108.28
Akaike Information Criterion (AIC): 22192.28
Bayesian Information Criterion (BIC):    22425.45

Phil Chalmers

unread,
May 7, 2017, 2:09:08 PM5/7/17
to Nikos Tsigilis, mirt-package
No idea how IRTPRO works. What happens when you analyse each group separately, as two independent single group models? 

Phil

--
You received this message because you are subscribed to the Google Groups "mirt-package" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mirt-package+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Nikos Tsigilis

unread,
May 7, 2017, 6:27:15 PM5/7/17
to mirt-package, ntsi...@city.academic.gr
Dear Phil,

I run seperate models for males and females with mirt and IRTPRO and I got identical results.  Findings are consistent with the multipleGroup function.  The discrepancy was caused  because IRTPRO by default fixs the mean and variance of the focal group to 0 and 1 respectively and frees the mean and variance of the reference group. When I set both groups' means and variances to 0 and 1 IRTPRO provided the same results with mirt. 

Thanks a lot

Νίκος

Phil

To unsubscribe from this group and stop receiving emails from it, send an email to mirt-package...@googlegroups.com.
Reply all
Reply to author
Forward
0 new messages