Natalie posted on Tuesday, March 02, 2010 - 2:06 pm
I know one should not use standardized scores as indicators of a latent growth curve. However, we have data for a variable that was collected across time that is standardized within children's sex and within classrooms. We have data for children in 30+ classrooms (30 at T1 and the number grows over time). Is it okay to use data that were standardized in this manner in either a regular latent growth curve or in a growth mixture model?
Bengt, to follow-up on your answer to this question. Could you please indicate why? Is this because when using standarized scores, that increases or decreases in mean scores cannot be found? Or is this for another reason? In some cases we are only interested in associated growth between growth parameters (variances, not means). Can standarized scores be used for that? And, is your answer to the above question specific for standardized scores within classrooms, or does this apply to any standardized scored (so also standardized over the population)?
The key question to ask is if we didn't do/hadn't done any standardization, would the growth model estimates come out the same or does the standardization distort the growth picture?
The example has 3-level data: time, person, classroom. And standardization within classroom. I would expect that to distort the picture. You can simulate data in Mplus and try it out.
Seltzer, Frank, Bryk (1994). The Metric Matters: The Sensitivity of Conclusions About Growth in Student Achievement to Choice of Metric. Educational Evaluation and Policy Analysis Spring 1994, Vol. 16, No. 1, pp. 41-49