Using Survey Software to Analyze Your Answers II
This post continues our discussion on how you can use sophisticated survey software to analyze the answers you receive.
In addition to giving you the choice of frequencies in various kinds of percentages when showing the answers to multiple-choice questions, more capable programs let you see various statistics. There are two kinds of statistics: descriptive and significance testing. Descriptive statistics describe your answers. The most common of these is probably the mean, sometimes called the average. This kind of statistic can be helpful when the numbers you use to code your answers are meaningful (no pun intended). If your question is “What is your gender?”, the numbers of the codes you assigned to men and women has no meaning. It’s just an identifier. But if your question is something such as “On a scale of 1 to 5, how would you rate this product?”, then the numbers do have meaning.
Significant statistics can be used both when the numbers do not have inherent meaning and when they do, you just need to use different ones for each case. When the numbers are arbitrary, such as in our gender example, you can use chi-square is or difference between proportions tests. The latter are sometimes called z-tests. The statistics tell you how likely any differences you find between groups in your data are reflect real differences in the population you are studying. For multiple-choice questions I too commonly used significant statistics are chi-square and difference between proportions tests (sometimes called Z-tests). Chi-squares give you a single number telling you how likely two variables are related. Difference between proportions tests can tell you how likely it is that different percentages of different groups of people who picked a particular choice is likely to reflect real differences in the population. The most capable programs let you choose whether you want to compare people and one group to everybody else combined or to other individual groups of people.
The most capable survey software also give you the ability to create summary tables. These tables combine the answers to two or more related questions. For example, you may have asked for ratings of five similar items on a 10 point scale. You should be able to create tables in which each row represents one of the items and can show you the mean, standard deviation and other descriptions of the answers given by different demographics. You should also be able to display statistics, such as t-tests between the different groups.
A similar format is when you ask questions such as “How many acres of corn, wheat and soybeans do you plant?” You should be able to have a row for each crop in which you can display the mean or average acreage of each crop, the total acreage of each crop and the percentages those totals reflect of the overall total number of acres planted.