When analyzing your survey, there are nearly infinite ways to look at your data. It may be helpful to think about different types of data in terms of how they can be measured to determine central tendencies of each. Measures of central tendency (mode, median and mean) attempt to describe a group of data with a single point. In our last article, we reviewed Median Measurement, now we’ll discuss Mean Measurement.
Mean Measurement: Data Types: Ratio (or Interval)
Mean is likely the measure of central tendency that is the most familiar and can be used to analyze ratio and interval data. While somewhat easier to analyze, these types are data are less-commonly used in survey data collection because most survey platforms do not handle this type of collection easily. Even asking simple questions about costs become challenging to collect in a way that can be analyzed by means because they require open-ended type questions that can be problematic for respondents to enter in a standard way. Also, many of the questions of interest lend themselves to this type of data. That said, the mean is just the average and can be easily calculated using:
If you are just looking at individual questions, you can consider just reporting the means. However, if you want to know if age or location affect responses, you may want to consider using T-tests or the Mann-Whitney test (de Winter & Dodou, 2010).
Weighted Means Measurement
Weighted Means Calculation:
If you don’t need to make comparisons between groups, but want to make comparisons between questions you can also calculate the weighted means of the questions as follows:
Since these were randomly generated data you can see there is no clear answer to which product people are more likely to purchase soon. Since these answers aren’t exclusive or ranked, there could be a tie. But you can also see that the average answer is right around the middle point, which is also expected with randomly-generated data. If you should see this pattern your participant-generated data, you should definitely take a closer look at the responses.
Rank order data:
A popular way to analyze rank order questions is to also use weighted means. It’s very similar to scoring the Likert-type questions above. Looking just at the means, and not taking into account any other type of analysis, it looks like people ranked sausage as their most preferred topping most often. You can again go back and a perform the Chi-Square (2) Goodness of Fit test to determine if the distribution is different than the expected values due to chance. Again, you may find that this distribution isn’t really different than the expected values from chance and in that case you may conclude that participants did not really have any clear preference for toppings or that participants answered randomly.
For the rank order analysis, it doesn’t matter whether you assign Rank 1 with the highest number or the lowest number. It is, however, important to be consistent within the question and remember whether those with higher weightings are ranked first or last.
In this article, we discussed how to analyze several types of closed-ended questions. However, there is often a need to collect more exploratory data that don’t limit respondents to a list of pre-defined answers. In coming weeks, we’ll discuss how to collect, code, clean and analyze free-response questions.
de Winter, J.C.F. and D. Dodou (2010), Five-Point Likert Items: t test versus Mann-Whitney-Wilcoxon, Practical Assessment, Research and Evaluation, 15(11)