BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:Cambridge Psychometrics Centre Seminars
SUMMARY:Effects of ignoring clustered data structures in f
actor analysis and item response theory - Dr. Jan
Stochl\, University of Cambridge
DTSTART;TZID=Europe/London:20120503T140000
DTEND;TZID=Europe/London:20120503T150000
UID:TALK37023AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/37023
DESCRIPTION:In research\, data for analysis come principally f
rom two sources: directly from the respondents the
mselves and from interviewers/raters. In the latte
r case\, clustering by interviewer/rater needs to
be considered when performing analyses such as fac
tor analysis or item response theory modelling (IR
T)\, although it is usually ignored. We use simula
ted data to study the consequences of aggregated a
nalysis (i.e.\, analysis ignoring clustering) on f
actor analytic estimates (both exploratory factor
analysis (EFA) and confirmatory factor analysis (C
FA)) and fit indices when the data are clustered.\
n\nOccasionally\, certain aspects of the hierarchi
cal information on clustering displayed by data ar
e partly known (as in the case of clustering by pa
tient service or treatment site\, for example). Ho
wever\, information about the interviewers within
each service is likely to be missing. In such case
s\, it might be better to consider using the avail
able information to improve the quality of factor
analytic estimates rather than completely ignoring
the hierarchical structure of the data. We study
the usefulness of this approach using simulated da
tasets. We also study the performance of different
estimators - maximum likelihood\, weighted least
squares and Markov chain Monte Carlo - on factor a
nalytic estimates when hierarchical clustering is
ignored.\n\nThe results show that ignoring cluster
ing in the data leads to serious underestimation o
f the factor loadings and item thresholds in ordin
al IRT treatment of rating data. In addition\, fit
indices tend to show a poor fit for the candidate
structural model. The Markov chain Monte Carlo (M
CMC) estimator shows better robustness than the ma
ximum likelihood and weighted least squares approa
ches. Partial information on clustering helps to c
orrect (and may overcorrect) fit indices\, but unf
ortunately\, it does not improve the factor analyt
ic model estimates themselves.
LOCATION:Seminar Room\, The Mond Building\, New Museums Sit
e
CONTACT:Luning Sun
END:VEVENT
END:VCALENDAR