- Subjective criteria
- Do not include trivial factors that have one and
only one of the original variables loading on the factor
- If the partial correlations among the variables (after
adjusting for the common factors ) is still large,
consider reducing the number of factors
- Objective criteria
- First a value is chosen at random for the number of
underlying factors
- Hypotheses testing is done as to the sufficiency of the
number of factors
- Statistics used are Likelihood Ratio Test (LRT) and
Akaike's Information Statistic (discussion is beyond the
scope of this talk)