Dear All
What is the restriction about the number of exogenous variables in accordance to the sample size. For my case, the number of observations in my panel data is 63. Is the number of 6 exogenous variables suitable?
Dear Imene,
I do not remember the material where I read it, in view of the number of observation and accompanying degree of freedom, and the variance distribution, the rule of thumb is 10 observations should support each explanatory variable. But, it is my feeling with this sample size it is very difficult to conduct an efficient estimation.
Hello Imene,
The short answer is, a bigger sample size is better!
The more elaborate answer hinges on how many parameters are proposed to exist (and therefore be estimated) in the model you will evaluate. Since you didn't specify your model, I can only talk in general terms.
The maximum number of parameters that could be evaluated in a given model would be k (k + 1) / 2, where k is the number of measured/observed variables. Since you have six exogenous variables, and, I presume, at least one endogenous variable, that means you could be estimating up to (7*8/2 = ) 28 parameters (7 individual variable variances, plus 21 possible covariances/correlations among variables). In a given model, you can fix some of the potential parameters to a specified value (e.g., 0, if you propose that two variables are unrelated to one another); doing so reduces the number of parameters to be estimated.
Here's one source (
core.ecu.edu/psyc/wuenschk/MV/SEM/SEM-Intro.doc) that, as Kidanemariam Gebregziabher suggests, recommends a lower bound of 10 cases per parameter to be estimated.
So, if you had a fully saturated/just identified model, with 28 parameters to be estimated, the minimum recommended size would be 280 cases.
Good luck with your work!
Thank you dear Kidanemariam and dear David for your relevant answers.
They really help
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size