Some Notes On Factor Analysis of Behavioral Data
In: BehaviourSearch for other papers by Robert Short in
Current site
Google Scholar
Search for other papers by John Horn in
Current site
Google Scholar
Abstract
The use of factor analysis with behavioral data is examined critically. Although there are good reasons to make wider use of factor analytic techniques in behavioral research, several important pitfalls should be avoided. Samples should be of adequate size, both absolutely and relative to the number of variables analysed. Variables should be reliable and devoid of scoring dependencies. One should not allow the same components to appear in different ratio variables, difference scores and composites. Variables and subjects should be carefully selected to represent a substantive design for the study. Ideally this design should stem from empirically-based, well reasoned theory. Careful thought should be given to the question of the number of factors to extract. Rotation of factors should be based on the model reflected in the design for the study. One should be wary of using mathematical-statistical indices with factor coefficients. The correlation, as such, and the standard error for a correlation, do not have the same properties when applied to arrays of factor coefficients as when used with free-to-vary numbers.
Purchase
Buy instant access (PDF download and unlimited online access):
Institutional Login
Log in with Open Athens, Shibboleth, or your institutional credentials
Personal login
Log in with your brill.com account
| All Time | Past 365 days | Past 30 Days | |
|---|---|---|---|
| Abstract Views | 285 | 62 | 3 |
| Full Text Views | 143 | 2 | 0 |
| PDF Views & Downloads | 22 | 2 | 0 |
Some Notes On Factor Analysis of Behavioral Data
In: BehaviourSearch for other papers by Robert Short in
Current site
Google Scholar
PubMed
Search for other papers by John Horn in
Current site
Google Scholar
PubMed
Abstract
The use of factor analysis with behavioral data is examined critically. Although there are good reasons to make wider use of factor analytic techniques in behavioral research, several important pitfalls should be avoided. Samples should be of adequate size, both absolutely and relative to the number of variables analysed. Variables should be reliable and devoid of scoring dependencies. One should not allow the same components to appear in different ratio variables, difference scores and composites. Variables and subjects should be carefully selected to represent a substantive design for the study. Ideally this design should stem from empirically-based, well reasoned theory. Careful thought should be given to the question of the number of factors to extract. Rotation of factors should be based on the model reflected in the design for the study. One should be wary of using mathematical-statistical indices with factor coefficients. The correlation, as such, and the standard error for a correlation, do not have the same properties when applied to arrays of factor coefficients as when used with free-to-vary numbers.
| All Time | Past 365 days | Past 30 Days | |
|---|---|---|---|
| Abstract Views | 285 | 62 | 3 |
| Full Text Views | 143 | 2 | 0 |
| PDF Views & Downloads | 22 | 2 | 0 |
