Using Survey Software to Analyze Your Answers III
Our previous post started our discussion of summary tables that combine the answers to multiple related questions. In the last post we talked about tables that we show that means of the answers given to a series of related questions, such as rating several items on a 10 point scale. Another type of summary table is a table that shows the distribution of the answers to a series of related questions. For example, you might have a series of questions asking people how much they agree with a series of statements. A table could have rows for each of the questions and columns for the answer choices. So each row would show you how many people agreed or disagreed with each statement and to what extent.
One limitation with summarizing the answers to a related series of questions in that format showing how many people pick each answer choice is that you cannot easily compare how different demographic groups answered the questions. Because the table could only have two dimensions, if the rows represent the different questions or statements and the columns represent different answers, that is the two dimensions available in the table. If you want to compare how men and women answered those questions in that format, you have to produce one table for men and another table for women and look at them.
Often it is more helpful to create a top box or top two box table. Suppose your agreement question offered five choices: strongly agree, agree, neither agree nor disagree, disagree and strongly disagree. Sophisticated survey software lets you create a table that summarizes how many people agreed with each statement, in other words how many picked either strongly agree or agree. Those could be combined into a single percentage so that you can then use the rows of the table to represent the different demographic groups. That way you can easily compare the total agreement with the statements from men, women, different age groups and so on.
This format would also let you use statistics to see if the different levels of agreement in different groups were statistically significant, bigger than could be expected due to chance to terms of the people who answered your survey. You can’t do this sort of comparison if you produce separate tables for each demographic group.
The most capable survey software offers additional ways to make your tables more attractive and easier to analyze. One such option is ranking. Suppose you have a question presenting a list of 10 magazines and asking people which ones they read. If you have your program show the magazines in popularity order, rather than the order you had them in the list, it will be easier to see which magazines are more popular. One wrinkle on ranking is one that separates the more limited programs and the more capable ones. Suppose in addition to those 10 magazines, you had a choice “none of the above.” Your program should be able to display the magazines in rank order, while always leaving none of the above at the bottom, even if it was picked by more people than some of the magazines in the list.
Another form of ranking option that is useful is to have your program automatically combine infrequently picks choices into a category such as “other.” Suppose you asked about 50 magazines, instead of 10. You would like to get many magazines that were only picked by a small number of people. You should be able to have your table show all the magazines that received less than a specified percentage of the answers grouped together. Doing so will make your table much easier to read.