Is there a lavaan example for before/after intervention data?

[Wang cs001632]

     For medical dataset, variable usually only contain before/after intervention time. Did two time point can be used to infer causal relationship?

[Mauricio Garnier-Villarreal]

The issue of "causal relationship" is more complex that that. If you only have pre and post intervention, you still cannot state causality. As a minimum you need a control group for comparison

Now, with that type of data I recomened the latent change score model as

mod_ch_tot <- '
tot_post ~ 1*tot_pre
change =~ 1*tot_post

change ~c(m1)*1
change ~~ c(v1)*change

tot_post ~~ 0*tot_post
tot_post ~0*1

change ~~ tot_pre

### cohen d
cd1 := m1/sqrt(v1)
'

[Nickname]

Wang,
  Shadish, Cook & Campbell (2002) discuss the design you describe as a Single-Group Pretest-Posttest Design.  They discuss strengths and limitations, and also conditions that moderate valid causal inference.  However, in general, this is a weak design.  See Chapter 4 for details.  Control groups offer one way to strengthen the design, but not the only way.  Shadish, Cook and Campbell assume that the researcher is not in possession of complete knowledge of all the causal relations relevant to the two variables of interest.  Much current work on causal inference in focused on exploring available inferences when such complete knowledge is available to the researcher.  Under such circumstances, the design can be sufficient for estimating causal effect sizes.  However, these methods do not apply if the research it in the phase of figuring out the qualitative causal structure.

Keith

Shadish, W. R., Cook, T. D. & Campbell, D. T. (2002).   Experimental and quasi-experimental designs for generalized causal inference.  Boston:  Houghton Mifflin.

------------------------
Keith A. Markus
John Jay College of Criminal Justice, CUNY
http://jjcweb.jjay.cuny.edu/kmarkus
Frontiers of Test Validity Theory: Measurement, Causation and Meaning.
http://www.routledge.com/books/details/9781841692203/

[Wang cs001632]

Thanks for your reply!

[Wang cs001632]

After searching internet for days, I find LCSM tutorial was rare.

I have find one, Growth Modeling_ Structural Equation and Multilevel Modeling Approaches .

Do you have other recommandation?

[Mauricio Garnier-Villarreal]

The book Growth Modeling: Structural Equation and Multilevel Modeling Approaches is the best resource I can think of. Their R example code is in OpenMx not lavaan

[Wang cs001632]

 Thanks. I have added more detail as the newest question on this website. Any suggestions will be helpful.

在 2019年9月7日星期六 UTC+8上午1:56:47，Mauricio Garnier-Villarreal写道：

[nick judd]

I highly recommend Keivit's article and scripts on OSF (https://osf.io/4bpmq/), Henk & Castro-Schilo 2016 along with McArdle's 2009 annual rev is good.

None of this is worth it, if you don't have multiple tasks probing the same construct at both timepoints (it's okay to have missing data). Mauricio's code is the equivalent of subtracting t2 from t1 and looking at the relationship of that to your groups (delta ~ group), there's no point to do that in lavaan unless you'll do more (be careful in your interpretation; i.e. reversion to the mean).

If you only have 1 task pre and post and your treatment was randomized consider just using residualized-change in a linear model (t2 ~ randomized_IV + t1).

I'm not an expert on causality, yet I remember reading issues when you only have 2 timepoints (e.g. pre, post). Another consideration is that you need to have at least strong longitudinal measurement invariance (see keivit's scripts for strict). 

Best,

Nick

Kievit, R., Brandmaier, A., Ziegler, G., van Harmelen, A.-L., de Mooij, S., Moutoussis, M., … Dolan, R. (2017). Developmental cognitive neuroscience using Latent Change Score models: A tutorial and applications (p. 110429). https://doi.org/10.1101/110429

Henk, C. M., & Castro-Schilo, L. (2016). Preliminary Detection of Relations Among Dynamic Processes With Two-Occasion Data. Structural Equation Modeling: A Multidisciplinary Journal, 23(2), 180–193.

Castro-Schilo, L., & Grimm, K. J. (2018). Using residualized change versus difference scores for longitudinal research. Journal of Social and Personal Relationships, 35(1), 32–58.

McArdle, J. J. (2009). Latent variable modeling of differences and changes with longitudinal data. Annual Review of Psychology, 60, 577–605.