BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Why latent varaibles in SEM do not always work well - Kasia Julia 
 Doniec
DTSTART:20141125T160000Z
DTEND:20141125T170000Z
UID:TALK56371@talks.cam.ac.uk
CONTACT:Chan Yin Wah Fiona
DESCRIPTION:Optimisation of 'measurement model' using expected and confirm
 ed SEM links as construct validity constraint.\n\nIn structural equation m
 odelling (SEM)\, the sets of links from latent variables (LVs) to their ma
 rker measures are often called ‘the measurement model’\, to distinguis
 h them from the regressions linking the construct variables (whether obser
 ved or latent)\, making up ‘the conceptual model’. This distinction ca
 n be useful under certain states of theoretical or methodological knowledg
 e\, and of availability of variables and data-sources\, including applicat
 ions in psychometrics such as testing between different models for covaria
 nce structures (somewhat similar to differing factor analytic solutions). 
  As the embodiment of a theory\, SEM provides a good framework for address
 ing general construct validity of measures. Thus\, the main deployment of 
 SEM outside psychometrics is for making medium-to-strong causal inferences
  from observational (ie non-intervention) data\; within this\, an LV that 
 assists goodness of fit has to both express both an efficient summary of c
 ovariance (like a principal component)\, and an assertion that the majorit
 y of the supposedly causal regressions in and out are similar for all obse
 rved variables marking the LV. But this is not always true. This second in
 fluence (or requirement for the LV to assist model fit) may or may not ass
 ist good measurement of a construct.  The ‘measurement model’ in psych
 ometrics can require more traditional and labour-intensive psychometric me
 thods for developing and improving measurement\, and showing that the meas
 ure is as good as it can reasonably be (eg by Pearson correlations). \n\n\
 nWe illustrate these points via our use of an adaptive strategy over a per
 iod in developing SEMs for two overlapping datasets on children’s middle
  ear disease and the consequences of this for development and quality of l
 ife. We used the first iteration of SEMs as general context for construct 
 validity and worthwhileness of the enterprise. We then returned to a secon
 d iteration of measurement (item selection\, scaling and weighting) and qu
 antified the improvements achieved. In some instances\, the re-scaling of 
 the score-values allocated to items’ response levels improved measuremen
 t\, as shown by enhanced regression coefficients between variables with al
 ready highly significant regression coefficients\; in others it left the r
 egression at least no worse\, but in revisiting we achieved better handlin
 g of missing data. We largely displaced the ‘measurement model’ into p
 rior regressions and principal component analyses with the aim of balancin
 g of validity with reliability and quality of distribution. In the ‘conc
 eptual model’ for causal paths to development and quality of life\, LV m
 odels were less successful than alternative parallel and serial structures
 .\n
LOCATION:2nd Floor Seminar Room\, Department of Psychology\, Downing Site\
 , Cambridge
END:VEVENT
END:VCALENDAR