Conducting an 86-variable Factor Analysis on a Small Computer and Preserving the Mean Substitution Option

Abstract

This paper shows how we overcame limitations imposed on us by the memory capacity of the relatively small mainframe we used in conducting a factor analysis in which means are substituted for missing values. Insufficient memory did not permit us to employ SPSSX, with its mean substitution feature, in conducting a factor analysis of 86 variables reflecting ways in which parents cope with the hospitalization of their children. Instead, we employed a two-step solution: (1) we ran SPSSX Condescriptive to create z-score equivalents of the 86 variables and recoded the z variables' system missing values to zeros; (2) the output of the Condescriptive run constituted the input of a BMDP P4M factor analysis run.