Decide on the variables to be used in the factor analysis. A rule of thumb given in "The Essentials of Factor Analysis" by Dennis Child states that you need least five data points per variable; use your sample size to find the most suitable number of variables. In short, divide your sample size by five to determine how many variables you will include. For example, if you have an original sample size of 50, you should include at most 10 variables in your study.
Find the number of factors. Run principal components analysis on your data. Plot the eigenvalues from the output of the principal components analysis against the principal components. Find the last principal component with an eigenvalue greater than one. The number labeling this principal component will be the number of factors for your factor analysis. If you see the third principal component's eigenvalue drop below one, then your number of factors will be three.
Apply the common factor model to the data. Initialize the communalities as the squared multiple correlations. The common factor model will yield a solution in terms of factors. You can treat these factors as new variables, representing a simplification of the original variables. As an example, if you included 10 variables and calculated three factors, you have reduced the full individual meanings of these 10 variables into three factors.
Interpret the factors. Apply multiple forms of rotation to the factors from the common factor model. Two common methods of rotation are Kaiser's varimax rotation and the quartimax rotation. Rotate your solution until you can interpret it easily in ordinary language. For example, if you were analyzing cars through factor analysis and you rotate your variables so that "luxury" and "quality" load onto the same factor, you may interpret this factor as "reputation." Likewise, if you find that "durability" and "performance" load onto the same factor, an interpretation of "practicality" would be suitable.