estimate item parameters with fixed abilities

[Tzur Karelitz]

For calibration purposes, we need to estimate item paramters while fixing the abilities on their known values which we obtained from a seperate analysis. 

How can this be done? it seems that the PARS method only fixes items, not people.

Thank you.

Tzur

[Guido]

I am not sure about this, but you could perhaps use the simdata function to simulate item responses based on your known item parameters and your (fixed abilities) theta matrix?

Then, use this object in the mirt function and specify 'customTheta'  - your custom Theta grid /fixed abilities in matrix form - , as a technical argument (see manual)?

PS:  I think you need to Import the original d-parameters, not b-values. In the mirt parameterization, the functions are written with the logistic metric, i.e., aiθj + di where di denotes item easiness. For the unidimensional models, the d parameters can be converted into traditional IRT b parameters. When IRTpars = TRUE, b = −d / a, while the a parameters will be identical under this parameterization.

# example a1 = 2.522040 and d = -2.1444262

# IRTpars = TRUE converts to a1 = 2.522040 and b = 0.8502744

# Hence, if you want to convert b-values into d-values than d = -1 * (0.8502744 * 2.522040) = -2.1444262

Again, I'm not sure about the above but I hope it gives you a starting point.

Kind Regards,

G.

[Tzur Karelitz]

Thank you, I will try this.