Page 1 of 1

factor analysis: variance accounted for

Posted: Thu Aug 04, 2011 12:31 am
by maxchen
Ver: EViews7Patch_072611.exe
When I replicated the example in factor analysis, the result confirms the output in page 724 of User guide II. However, for the summary information on the total variance and proportion of common variance, I am puzzled with the resuld
1 the "Cumulative" column, how to explain the total cumulative? see the 3rd line of the text, the cumulative variance accounted for by the two factors is 3.546949, not 6.27 (6.266613), only 50% (not 90%) of the total variance.
2 the "Proportion" column, gives the proportion within the factors, not for total variance from observed variance/correlation matrix. It is much informative to report the proportion of total sample variance due to each of the factors

Re: factor analysis: variance accounted for

Posted: Thu Aug 04, 2011 9:31 am
by EViews Glenn
It is customary to report the fraction of the common variance, not the total. This is well defined since the factors are orthogonal.

Note that interest here isn't in a measure that is R2-like, but rather in what part of that which we explain we can attribute to a given factor. There is a whole set of goodness-of-fit measures that serve the former purpose.

I think you're right about the text and total cumulative being off. I'll make a note of this.