What Is Cross-Tabulation Analysis?
Cross-tabulation analysis is a market-research tool that aims to show the relationship -- or lack thereof -- between certain preset variables. For example, if you surveyed 1,000 people about their favorite breakfast cereal, you could create a table cross-tabulating cereal choices with the age bracket of the people you surveyed, to see how age may affect breakfast preference. Cross-tabulation for large data sets is easier when it's done on a computer.
Other People Are Reading
-
Options
-
You don't have to stop at one cross-tabulation table -- create as many tables as there are relationships between variables that you want to investigate. In addition to tabulating cereal preferences by age, for instance, you can also tabulate selections by income, race, geography and level of education. The only limitation is that you collected data on the variables in your original survey. Cross-tabulating data may show that variables are strongly correlated, but it sometimes shows they have no actual relationship.
Chi-Square
-
Even if you think you see a relationship between variables, it may be a fluke. Chi-square testing is a mathematical method that compares the results of cross-tabulation to those you would observe if the results were completely random, and the two variables didn't affect each other. Several software programs on the market as of this publication can handle the number crunching involved. This reduces the work involved in analyzing large surveys with many variables to cross-tabulate.
-
Hypotheses
-
A computer can crunch numbers, print up tables and calculate the chi-square, but it can't tell you what information is important to your project. Before you gather data, formulate a hypothesis you want to test -- kids like sugary cereal more than adults do, for instance -- then make sure the survey collects the information you need to confirm or reject the hypothesis. Don't commit yourself to an unproven hypothesis: If the information shows it's wrong, you need to accept that. [ref3
Caution
-
Exercise care when you're drawing conclusions from cross-tabulation. Even if the computer shows a very strong link between age and breakfast tastes, that may not mean much if you have only a half-dozen survey respondents under the age of 12. Small numbers are more vulnerable to sampling flukes, such as that you just happened to survey six children who share the same taste; a larger sample in such a case might cross-tabulate differently. This is an example of how analyzing the computer's information requires the use of judgment, not just statistics.
-
