Factor analysis primer: make sense of complex survey data

What single measure describes your overall health?

Your blood pressure? Calorie intake? Weight?

None of these numbers gives you the full picture by themselves, but putting them all together can tell you a lot.

This is an example of when a researcher might use factor analysis, a statistical technique that’s useful for simplifying and analyzing large sets of data with many variables.

It can help you find out whether variables (or in the case of surveys, questions) are correlated with one another or with some other variable or concept. Statisticians call these related variables common factors.

You can determine the relationships between groups of variables by lumping together the ones that are that are strongly correlated, making them into common factors. This is the basis of factor analysis, which is often used in the fields of psychology, health, and political science.

For example, take these questions you might find in a healthcare survey.

Individually, the responses to these questions are too specific to say much about your overall health. But taken together, they can provide a more comprehensive measure of your wellbeing, which is the common factor that researchers are truly interested in.

If a researcher asked just one broad question about your health, say, asking you to rate your overall health as excellent, very good, good, fair, or poor, you might have a hard time answering.

After all, are you comparing yourself to your elderly neighbor? Your teenage cross-country runner daughter? It’s hard to know where you stand in relation to everyone else, and different respondents might compare themselves to different groups.

Instead, many researchers will ask a series of health-related questions and perform a factor analysis, which generates a standardized score of health.

There are three primary uses for factor analysis:

Note: Factor analysis is an advanced technique that requires a statistical software package. You should be very familiar with one of these packages before you begin your work.

Now that you know what factor analysis is, here are some survey design tips you’ll want to keep in mind if you plan on using factor analysis:

Factor analysis relies on having lots of data. Even if you’re using a sample size calculator, the exact number of respondents required to do a factor analysis will depend on things like your population size and the questions you’re asking, but the more completed responses you have, the better.

Factor analysis allows you to summarize broad concepts that are hard to measure by using a series of questions that are easier to measure. The idea is to gather a lot of data points and then consolidate them into useful information.

You need quantitative data in order for factor analysis to work, so the answer options to your questions should fall on a scale. It doesn’t matter whether you are using a number scale (e.g., from 0 to 10), a binary scale (e.g., Yes or No), or a Likert scale (e.g., strongly agree/agree/neutral/disagree/strongly disagree). The only requirement is that your options should be ordered in some way.

Plenty of analysis—generating charts, graphs, and summary statistics—can be done inside SurveyMonkey’s Analyze tool. That means the majority of SurveyMonkey customers will be able to do all their data collection and analysis without outside help. But factor analysis is a more advanced analysis technique.

If you are already comfortable working with statistical software packages like R, SAS, SPSS, or Stata, just export your survey data from Analyze to download the data into the format that fits your software.

While casual survey makers might not have the need (or software) for the level of detail factor analysis provides, it can be an invaluable tool in a survey researcher or statistician’s kit. By boiling down multiple data points into digestible chunks, you can measure concepts that are otherwise difficult to quantify and spot relationships in your data.