[IPAC-List] Answer to MULTILOG Question

[Christopher Cerasoli]

Hello All,

 

Thank you for your responses and feedback.  Based on feedback both from the
IPAC listserv and the helpful customer service at SSI, I've been able to get
an answer to my question.

 

Despite its options to run multiple tests with one syntax in the Graded
Model, MULTILOG can NOT perform tests for multiple abilities/thetas with one
syntax run.  The option to run multiple tests is only if you wish to specify
different clusters of items for the same test.  Each MULTILOG Graded Model
run must be (and is considered as) an overall test of a unidimensional
construct.

 

Regards,

 

Chris

 

~~~~~~~~~~~~~~~~~~~~~~~

Christopher P. Cerasoli

Graduate Assistant

State University of New York

University at Albany

Department of Psychology

Social Science 375

1400 Washington Ave.

Albany, NY 12222

cc572532 at albany.edu

~~~~~~~~~~~~~~~~~~~~~~~

 

 

-----Original Message-----
From: Christopher Cerasoli [mailto:cerasolic at gmail.com] 
Sent: Sunday, October 10, 2010 2:25 PM
To: ipac-list at ipacweb.org
Subject: [IPAC-List] MULTILOG Question

 

My question refers to the MULTILOG program for Item Response Theory data.

MULTILOG is a Window's based program that allows one to estimate item and
theta/ability parameters using different models under IRT (e.g., 1PL, 2PL,
Graded, etc.)

 

 

Using the Graded model, MULTILOG allows you to specify multiple tests, which
is useful if you want to estimate different abilities/thetas in one set of
data.  For example, you could estimate math capabilities, reading
capabilities, and science proficiency without having to enter three separate
datasets individually.  Or, you might have a personality inventory with five
parent constructs (e.g., conscientiousness, extraversion, etc.) and wish to
have MULTILOG analyze one dataset containing all responses.

 

 

The problem I run into is that while MULTILOG allows you to specify
different tests, with one dataset, the output suggests that it's really only
running one test.  I've considered the possibility of common method
variance, but can't imagine that would lead to such striking differences.

 

 

Has anyone else experienced this issue?

 

 

Thank you in advance,

 

 

Chris

 

 

~~~~~~~~~~~~~~~~~~~~~~~

Christopher P. Cerasoli

Graduate Assistant

State University of New York

University at Albany

Department of Psychology

Social Science 375

1400 Washington Ave.

Albany, NY 12222

< <mailto:cc572532 at albany.edu> mailto:cc572532 at albany.edu>
<mailto:cc572532 at albany.edu> cc572532 at albany.edu ~~~~~~~~~~~~~~~~~~~~~~~