Factors, then, are composed of several measured variables. The strength of the relationship between each measured variable and the factor that contains it is assessed by a statistic called a factor loading. Factor loadings are easy to interpret because they are interpreted the same as standardized regression coefficients. Statistical software will produce a table with each factor forming a column, and each variable in rows. Each variable will have a factor loading on each factor, which is the value in the cell where the variables intersect the factors. In the professional literature, factor loadings below a certain cut point are often omitted, making the table easier to interpret. Also, the variables may be ordered such that factors are easily identified. That is, the variables with high loadings on a particular factor are clustered together.
So far, we have considered factor analysis as a method of examining underlying factors. It can also be used to evaluate individual variables. Applied researchers can test the ability of one variable to serve as a proxy for several. In other words, if an individual item correlates very highly with a particular factor, then that variable can be used as a “stand in” for the more complex, harder to measure variable. Factor analysis can also be used to see if a particular item belongs with the factor that the researcher believes it does. The validity and reliability of standardized instruments can be enhanced in this way.
Last Modified: 02/14/2019