Hi everyone,
If I tried to run a structural equation model with three related dependent variables (one ordinal, one binomial, and one continuous), are 71 observations sufficient? I want to assess the effect of two categorical independent variables on the dependent variables, and would like to, ideally, control for experimental day (8 days of experiments, with varying number of trials per day).
Thanks so much!
Aliza
You could read Boomsma, A. (1983). On the robustness of LISREL (maximum likelihood estimation) against small sample size and non-normality. Amsterdam: Sociometric Research Foundation. (Doctoral dissertation, University of Groningen) (open access: http://aboomsma.webhosting.rug.nl/boomsma.htm). Or the revisited version: Boomsma, A., & Hoogland, J.J. (2001). The robustness of LISREL modeling revisited. In R. Cudeck, S. du Toit & D. Sörbom (Eds.), Structural equation models: Present and future. A Festschrift in honor of Karl Jöreskog (pp. 139–168). Lincolnwood, IL: Scientific Software International. Form SEM models one would want at least 200 cases. As you are using ordinal, binomial and continuous data, it will be more, but it will also depend on the complexity of the model you want to create. As you have mixed data, you might need a polyserial correlation matrix. As problem with polyserial matrices: they can be not positive semi-definite.
You can calculate power to determine sample size, but 71 is adequate
Seems to me 71 can not be adequate. Lindeman et al Introduction to bivariate and mulitvariate analysis (1980, Scott, Foresman and Company, pg 55) gives the advice 'Although larger samples are desirable, we shall recommend n=100 as a minimally acceptable sample size for regression and correlation studies'. Regression or correlation studies use interval data and are less complex then the model Aliza le Roux is trying to estimate. Boomsma et al show 200 is needed when MLE is used, which is a simple estimation procedure in SEM.
Thanks for the feedback, Peter Moorer and Andrew Paul McKenzie Pegman -- I appreciate it.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size