

Standard Errors, small sample 

Message/Author 


I've found differences in the standard errors reported in Mplus compared to SPSS, when doing exercesises with alumns about modelling a multiple regression with a small sample (50 cases) with the ML estimator. I'm quite sure there is some reason for this, but I don't know it. Best regards, Fernando. PS: The exercise involved selecting a random sample of 50 cases from the hsb2.sav file and then doing the regression (anova) of write ON female prog1 prog2. The estimates were exact, but the standard errors show differences even in the first decimal. 


I think this is most likely due to Mplus using n and SPSS using n1. 


Linda, thank you very much for your quick and helping answer. I understand that if I multiply the Mplus Standard Errors by (n/n1)^2, the result are the Standard Errors reported by general statistical packages (SPSS, Stata, etc.). Is that true? Thank you very much, Fernando. 


If the reason the standard errors differ is due to the difference in n versus n1, then if you multiply the Mplus standard error by n and then divide it by n1, they should be the same. If this is not the case, then other reasons they are different could be different sample sizes, different estimators, etc. If you want a definitive answer, you should send your Mplus input, data, and output, your SPSS output, and your license number to support@statmodel.com. 


Thank you. I send the requested data. Fernando. 

Back to top 

