How to Factor Survey Results

Factor analysis

Factoring survey results uses analysis techniques to reduce and analyze survey data to discover underlying influences, or factors, that may affect responses to questions. The surveys themselves do not measure these factors; this the reason they are considered underlying. Researchers must extract and identify the factors thought to influence respondents' answers to survey questions, based on a statistical analysis of the responses. Social scientists and market researchers use factor analysis to make sense of their survey data. Factor analysis involves complex statistical techniques and requires you to use a computer program capable of such analyses.

Instructions

1. Enter your survey results, numerically coding your responses, into a spreadsheet for the statistical software program you are using. Popular statistical packages include SAS and SPSS. You can use Excel, part of the Microsoft Office software package, but only if you download and install a supplemental statistical program, such as XLStat (see Resources 2), that expands Excel's statistical capabilities. The data analysis tool included in Excel is not designed for factor analysis.

2. Familiarize yourself with your survey data by examining the range of responses to your survey or questionnaire. For example, if your questions used a scale of response categories that asked respondents to indicate the extent to which they agree or disagree with each question or statement, see what percentage of respondents agreed or disagreed with each question. The better you know your survey data, the more informed your factor analysis will be.

3. Extract the number of factors in your survey data by calculating the correlations among separate questions, or items. Correlation measures (coefficients) range in value from minus 1 (perfect negative correlation) to plus 1 (perfect positive correlation). A coefficient of zero indicates no correlation. Factor analysis assumes that highly correlated measures (positive or negative) are likely influenced by the same factors.

4. Identify the number of factors influencing your respondents' answers to your survey questions by examining the eigenvalues, which are statistical measures of the amount of covariance (the extent to which items, or variables, change together). A common criterion in factor analysis is that the number of factors influencing responses is equal to the number of eigenvalues greater than 1. If your analysis shows three eigenvalues greater than 1, then three underlying factors influence respondents' answers to the survey questions.

5. Rotate your factors, using the factor rotation procedure in your statistical software program. This will produce a display that shows the correlations between each survey item and the underlying factors. Based on the patterns of correlation, you can then name and label these underlying factors. These factors then provide an explanation for how subjects respond to the questions posed in your survey.

References

• Stat Help: Overview of Factor Analysis

Resources

• Cornell University: Introduction to Factor Analysis
• XLStat: Statistical Program for Excel