Title
irt_me -- Calculates marginal effects for the latent variable (Theta) after IRT
models
General syntax
irt_me [varlist] , [options]
Overview
irt_me calculates marginal effects for the latent variable (theta) after an IRT
model. The latent variable is the independent variable; the variables specifed in the
varlist are the observed items which are the dependent variables in an IRT model. If
no varlist is specified, irt_me calculates marginal effects for theta across all of
the observed items.
irt_me supports all models that can be estimated using the Stata irt command and also
models for continuous and count items (regress, poisson, and nbreg). A mix of
different item types is also allowed (e.g., a mix of binary and ordinal items).
+---------------------------------+
----+ Required Option for gsem models +----------------------------------------------
latent(string)
is required if gsem was used to fit the model of interest. This is the
name of the latent variable you wish to obtain marginal effects for. The
latent( ) option should not be used if the model was fit with the irt
command as irt automatically names the latent variable Theta.
+---------+
----+ Options +----------------------------------------------------------------------
model(string)
specifies the name of saved model estimates to use. See estimates store
for saving model estimates. By default, irt_me will use the IRT/GSEM
estimates in memory. If the relevant model estimates are not in memory,
you must specify their name.
decimals(#) changes the number of decimal places reported for the statistics. The
default is 3. Any integer between 1 - 8 is allowed.
start(#) specifies the starting value for the prediction used in the calculation
of the marginal effect. The default is -0.5 for a default marginal effect
estimate of a centered +1 unit change.
end(#) specifies the ending value for the prediction used in the calculation of
the marginal effect. The default is 0.5 for a default marginal effect
estimate of a centered +1 unit change.
range calculates marginal effects across the trimmed range of the predicted
values of the latent variable theta. Predictions are made at the 1st
percentile of theta (start) and at the 99th percentile of theta (end)
title(string)
changes the title of the table of results. A default title is
automatically included.
help prints footnotes below the table describing what the columns in the table
represent.
Examples
webuse masc1
irt 2pl q1 q2 q3 q4 q5
irt_me, help
gsem (Theta -> q1 q2 q3 q4 q5, logit), var(Theta@1)
irt_me, latent(Theta)
Authorship
irt_me is written by Trenton D Mize (Department of Sociology & Advanced
Methodologies, Purdue University). Questions can be sent to tmize@purdue.edu