Complex with cluster data (pairs)...
Message/Author
 Jesus Camacho-Morles posted on Tuesday, May 23, 2017 - 1:29 pm
I have 200 participants who worked in pairs to complete collaborative problem-solving (CPS) tasks. I am studying the relationship between enjoyment and CPS performance. I believe this relationship is mediated by effort regulation.
So, I have two latent variables (enjoyment, effort regulation), one continuous dependent variable (CPS performance), and I would like to control for Gender and Prior knowledge in numeracy.
My data is nested, and participants were paired based on gender. So, I have 100 pairs, 50 pairs females, 50 pairs males

Is it correct to do SEM, Type=Complex, Cluster=pairs, to study this mediation?

Is it a problem that groups are comprised of only two people?

Any suggestions will be much appreciated.

Many thanks

Jesus
 Bengt O. Muthen posted on Tuesday, May 23, 2017 - 5:48 pm
Do you really have variation within the pairs for your variables?
 Jesus Camacho-Morles posted on Tuesday, May 23, 2017 - 5:54 pm
Hello,

 Bengt O. Muthen posted on Wednesday, May 24, 2017 - 4:33 pm
Use analysis Type = twolevel basic; with cluster=pairs.
 Jesus Camacho-Morles posted on Wednesday, May 24, 2017 - 5:13 pm
Thanks. I already run the analysis using type=complex; cluster=pairs. Do you think that type=twolevel basic is more appropriate?
I hope I don't have to do it all again
 Bengt O. Muthen posted on Wednesday, May 24, 2017 - 6:03 pm
I was suggesting it so you can see how much within-pair variance you have. It doesn't mean you have to use twolevel modeling.
 Jesus Camacho-Morles posted on Wednesday, May 24, 2017 - 6:51 pm
Ok, I see.
This is what I got:

Within level variance = 0.284
Between level variance = 0.465
 Bengt O. Muthen posted on Thursday, May 25, 2017 - 4:12 pm
Should be fine to do Type=Complex with your 2 members per cluster.
 Jesus Camacho-Morles posted on Thursday, May 25, 2017 - 5:18 pm
great! Thanks a lot