Ωnyx › Foren › General SEM Discussion › Second-Order Multiple-Group Latent Curve Modeling
Schlagwörter: Pre-test post-test control group, SEM
- Dieses Thema hat 2 Antworten sowie 2 Teilnehmer und wurde zuletzt vor vor 6 Jahren, 2 Monaten von admin aktualisiert.
-
AutorBeiträge
-
September 17, 2018 um 9:19 am Uhr #844adminAdministrator
Hi there. Before posing my question, I’d just like to say how great this software is; without it, my Neuromarketing Master’s thesis would be highly challenging – so, thank you!
I am conducting a study, adopting a pre-test and post-test control group design, with two (2) observations made both pre-test and post-test (i.e. two surveys and two behavioural tasks, one of each before and after treatment exposure). Both observations will be measured on a 0-200 scale and I would like to use SEM to assess the efficacy of the intervention (treatment), in particular Second-Order Multiple-Group Latent Curve Modeling (SO-MG-LCM).
I am focusing on the effectiveness of health interventions on both explicit and implicit attitudes of consumers toward foods. The survey (pre- and post-test) is used to measure (on a 0-200 scale) the explicit attitudes, while an IAT task (scores adapted to be measured on a 0-200 scale) is used to gauge implicit attitudes. While I will be testing for mediation effects of one measure on the influence of the intervention on the other measure, and vice versa, would it be recommended that I isolate each of these measures for the purposes of the SEM analysis (i.e. all explicit measures — both pre- and post-test and for both experimental and control groups — and all implicit measures are analysed independently)?
With the SO-MG-LCM approach, how exactly would I structure the model(s) in Onyx? Additionally, what values would I need to import into Onyx in order to test the model(s) (i.e. covariance, residual variance etc.)? I am an absolute beginner when it comes to SEM, so all feedback would be greatly appreciated! Thanks in advance
September 17, 2018 um 9:21 am Uhr #845adminAdministrator
Please click on this post to open a graphic representation of SO-MG-LCM.Figure 1. Second Order Latent Curve Models with parallel indicators (i.e., residual variances of observed indicators are equal within the same latent variable: ε1 within η1and ε2 within η2). All the intercepts of the observed indicators (Y) and endogenous latent variables (η) are fixed to 0 (not reported in figure). In model A, the residual variances of η1 and η2 (ζ1 and ζ2, respectively) are freely estimated, whereas in Model B they are fixed to 0. ξ1, intercept; ξ2, slope; κ1, mean of intercept; κ2, mean of slope; ϕ1, variance of intercept; ϕ2, variance of slope; ϕ12, covariance between intercept and slope; η1, latent variable at T1; η2, latent variable at T2; Y, observed indicator of η; ε, residual variance/covariance of observed indicators.
September 23, 2018 um 2:33 pm Uhr #846adminAdministratorHi Kappers,
great you’re using Onyx!
In the model figure you sent, there is just one variable (the eta) which are measured by two indicators, however, if I read your text correctly, you have two independent measures which don’t combine to a common factor score, is that correct? Or do you want to measure „attitudes“ as a common factor?
If not, I would first set up the model as it is given in your representation with both variables separately, and then use the other variable (at pretest) to predict the change by a regression path. For example, you could set up the model for explicit attitudes, and then add the implicit attitudes at pretest as a predictor variable and connect it to the eta_2, the latent change, to see how strong implicit attitudes predict the change in explicit attitudes. For the other direction, you could then set up the model with implicit and explicit attitudes exchanged. This is just a very quick suggestion from what I understood, I hope it helps a little.
Cheers,
Timo
-
AutorBeiträge
- Das Forum „General SEM Discussion“ ist für neue Themen und Antworten geschlossen.