Message/Author |
|
|
Hi, I have a data analysis question. I have three waves of data that is clustered by cohort and gender as a covariate. Example 6.1 seems to be my best option, but does not show with clustered data. Is there another example I should use or how might I run that analysis with the clustered data and covariate? Thanks! |
|
|
Clustered data analysis examples are shown in Chapter 9 of the UG. You say clustered by cohort. I don't know what that means because cohorts are usually not viewed as clusters. For a multiple-cohort example, see ex 6.18. |
|
|
Hi - sorry for the confusion. I think I used cohort incorrectly. I have three time points of data for a group based intervention and would like to test if the intervention led to change while accounting for group membership (I think similar to accounting for the fact that students are in classrooms). What might be the best example to get me started? Thanks! |
|
|
So are you saying that you have a situation like one Treatment group and one Control group and everybody is measured 3 times? If so, see Multiple Population Growth Modeling in the Topic 4 Short Course Video and Handout on our website. See also the paper: Muthén, B. & Curran, P. (1997). General longitudinal modeling of individual differences in experimental designs: a latent variable framework for analysis and power estimation. Psychological Methods, 2, 371-402. |
|
|
HI - actually there is only a treatment group in the sample, measured at 3 time points (is from clinical archival data), would that still be the appropriate technique? Thanks! |
|
|
So you want to simply study change in your group single group of subjects taking into account that observations are not independent across time due to being from the same subjects. If so, see the UG chapter 6 growth modeling examples - and also the corresponding Short Course Topic 3 video and handout on our website. |
|
|
OK - I think I figured it out, just want to double check. So I have 5 variables: time 1, 2, and 3 total score, the number of sessions the person attended of the intervention (numsess), and the group they were in (group). As I understand it, group and numsess are time-invariant. So my code should look like: DATA: FILE IS GrowUse.dat; VARIABLE: NAMES ARE t1_total t2_total t3_total group numsess; MODEL: i s | t1_total@0 t2_total@1 t3_total@2 ; i s ON group numsess; Is this correct? My big question is whether this properly accounts for the multilevel nature of the data. That is, is including group as a covariate the appropriate technique, or is there some other way I should account for this between subjects variable? Thanks! |
|
|
Yes, this is correct. As we discuss in our Short Course Topic 3, the multilevel nature of the data is taken care of in a wide, single-level format where correlations across time are captured by growth factors. |
|
Back to top |