[Rasch] GPCM Reliability Estimate

Mike Linacre mike at winsteps.com
Fri Nov 15 07:33:43 EST 2013

Thank you for your question, Randy.

1) compute the Person Estimates and each of their S.E.s using the GPCM 
parameter estimates. Hopefully these are reported by the GPCM software.
2) square and average the person S.E.s = MSE = Error Variance
3) Person "Test" Reliability = True Variance / Observed Variance
= (Observed Variance - Error Variance) / Observed Variance = (Person 
S.D.**2 - MSE) / Person S.D.**2


Mike L.

On 11/15/2013 5:58 AM, Shelley & Randy MacIntosh wrote:
> A colleague has been asked to compute an IRT reliability estimate for GPCM.
> I have been unable to find a source on that.
> Can anyone on this list suggest how we might go about this?
> Thanks,
> Randy MacIntosh

More information about the Rasch mailing list