Test for ignorable / equivalence.

Hope this helps

Matt

Hi,

in my POV, it depends on your conceptualization of the group constructs. Analogous to the distinction between reflective measurement vs. aggregation/composite building on the individual level, group level constructs can be thought as having an existence status independent from the contributing individuals. In this case, asking individuals bout the group climate simply results in a reflection of the group climate. A requirement of such a conceptualization would be high agreement among all individuals.

In contrast, the group construct may simply be the sum/aggregate of its individuals. For instance, "group performance" could be simply be the sum of the individual performances. In such a case, you don't need agreement (i.e., individual may have low performance whereas individual B has high performance). The construct is merely the sum of its parts.

See

Chan, D. (1998). Functional relations among constructs in the same content domain at different levels of analysis: A typology of composition models. Journal of Applied Psychology, 83(2), 234-246.

Stapleton, L. M., Yang, J. S., & Hancock, G. R. (2016). Construct meaning in multilevel settings. Journal of Educational and Behavioral Statistics, 41(5), 481-520. doi:10.3102/1076998616646200

In your case, adhering to scenario A would allow to model longitudinal changes even if the set of individuals differ. Even in the extreme case, were the set of T1 is not the same as at T2, you get a valid measurement of the group phenomenon. Of course, you would have to provide evidence (e.g., ICC) that this is reasonable.

In addition, as Matthew Kerry noted, you could include validation measures in a longitudinal (or group) model in which you test for equality of important parameters (e.g., relationships with these validation criteria that should not sign. differ across the time points.

Best,

Holger

It is relatively unusual to have both "panel" data where the same respondents are followed over time, combined with dat where different respondents answer the same questions over time.

Can you say something about how large your sample is and what proportion were present at both T1 and T2?

First of all, thank you to all respondents for your precious contributions!

I read about ignorable/equivalence tests and, if I understand their purposes, with these instruments, I would try to demonstrate that samples in T1 and T2 are "psichometrically" similar, right?

It is interesting and it helps me, as interesting are the comments and links you all provide. On the other hand, I ask if there may be a sense in testing relationships that, as expressions of climates, describe "the way the wind is", the "mood" in working centers/organization.

In other words, I wonder if there may be a theoretical reason (beyond the psychometric aspects), and some precedents in literature you know, to test relationships not caring about who is in T1 and in T2.

I think the concept of climate in an organization or in some departments may go beyond the stability of its staff, and for this reason I have this doubt.

Size of samples are quite large. Around 800 for T1 and 600 for T2 and, however, a substantial number of people is costant between the two surveys.

I precise that the questions of the surveys, when not posed as individual questions for individual constructs, have working centers as reference.