Abstract: Confirmatory factor analysis (CFA) has been widely used in psychological andeducational research. By summarizing and referring to results of three simulationstudies, the present paper discussed strategies panicularly relevant to small samplestudies. Specifically, it was suggested that more items should be used and factorloadings can be restrained to be equal within the same factor. Generally, larger sampleis more desirable, but when resource is limited, the above two strategies may solvenon-convergence problems and increase the estimation accuracy of important parameters.Furthemore, the use of two indicators within a factor, the use of less than fourindicators when sample size is only about 100, and the combination of items intoparcels and using their averages as indicators are not recommended because they maylead to non--convergence, large standard error and biased parameter estimation.

Key words: statural equation modelling, confirmatory factor analysis, multivariareanalysis, small sample