The previous Part 1 showed an example of aggregating various types of responses (single response, multiple responses, scores) in questionnaire data at once using the "categorical" platform.
That said, if it's just a summary level, you can do it with spreadsheet software, and if you're using a questionnaire system, the system will display the summary results in tables and graphs.
So what are the benefits of using JMP's Categorical?
It introduces this time Compare response distributions across groups That's what it means.
Specify grouping variable as categorical
In the survey analysis, in addition to summing up individual responses, responses are also aggregated by dividing them by respondent attributes (gender, age, etc.). This is the so-called cross-tabulation.
Consider the SQC training example covered in Part 1. In this example, there is an attribute "position" (general, with position), so let's aggregate and compare other answers by position.
Consider the following analysis of the data table above.
Position x Motivation for taking the course (multiple answers): Multiple responses
Job title × Experience using statistical software (single answer): Nominal scale
Position × time allocation, text structure, practicality (score): ordinal scale
Position x satisfaction (score): ordinal scale
At this time, in "Categorical", "Title" is changed as follows. [X, grouping category] to.
Proportion homogeneity test (do the proportions of responses differ between groups?)
First, let's take a look at the report for the multiple response "Why did you take the course" and the nominal scale "Experience with statistical software" (yes/no). As shown below, a cross-tabulation table that is divided into "general" and "with title" is displayed, and the corresponding share chart is also displayed.
Comparing the share of answers for "Why did you take the course" between general and job titles, there seems to be a difference in share in the following items.
Instructions from superiors: general (24.7%), with title (16.8%)
Interest in SQC: general (8.9%), with title (16.1%)
Regarding experience with statistical software, 39.6% of those with general experience and 50.9% of those with managerial experience also seem to have a difference.
What I'm saying here is that "there is likely to be a difference". I'm using statistical software, so I'd like to find out if there is a statistical difference. Therefore, use the test options under "Categorical".
Homogeneity test for non-multiple responses
For responses that are not multiple responses, as in the experience of statistical software, [Test for Homogeneity of Responses] Execute options.
Homogeneity of responses is tested across groups, and here we are testing the hypothesis that proportions of responses are equal across job titles. It corresponds to the chi-square test used to examine the independence of crosstabulations.
Since the p-values for the report likelihood ratio chi-square and Pearson's chi-square are less than 0.05, the 5% significance level of the test indicates that the responses are not homogeneous, i. indicates that there is
Homogeneity test for multiple responses
For the multiple-answer question "Why did you take the course?" [Multiple response test] is used. Here we assumed a selected/unselected binomial distribution for the response categories [Binomial distribution homogeneity test] shows a report of
The report, Binomial Test for Each Response, provides statistics comparing the proportions of responses for each question across groups.
For example, the total number and respondents for the response of "boss instructions" are as follows.
In general, 144 out of 392 (36.7% = 144/392)
26 out of 111 people with titles (23.4% = 26/111)
At this time, the chi-square value and p-value (=0.0074) are obtained by a logistic model where Y is the response rate (assuming a binomial distribution) and X is general/with a position.
Referring to the report "Binomial test for each response", significant differences are recognized for the two items "supervisor's instructions" and "interest in SQC". This suggests that general employees took the seminar because their superiors told them to take it, and that those with job titles were interested in SQC itself.
For other items, although there is a difference in the ratio between general and job titles, it does not seem to be a significant difference.
compare scores
In many cases, questions that ask about product satisfaction are answered with a score such as a 5-point evaluation. Then you want to see if there is a difference in scores between groups.
In this example, we asked respondents to rate the seminar time allocation, text structure, and practicality on a 5-point scale, and on their satisfaction with the seminar on a 10-point scale. If you want to compare these evaluation scores between job titles: [Score average comparison] I'll try using
This is a report on seminar time allocation, text composition, and practicality. The average and standard deviation when responses 1 to 5 are considered as scores are displayed on the right side of the cross tabulation table.
Looking at the last column, "Comparison of averages," you can see the letter "f" in row 5 (score for general utility, labeled E). This indicates that it is significant at the significance level of 0.1 for row 6 labeled F (score for practicality with title). (Please refer to the explanation surrounded by the red frame in the figure.)
We show that the average score for practicality is 3.52 for general and 3.37 for with position, but the p-value of the test (using the t-test) was less than 0.1 when these means were pairwise compared. Those with titles may have been a little dissatisfied with the practicality of the seminar compared to the general public.
Let's also refer to the report on the "satisfaction" of the seminar.
The general average is 7.23, the title average is 6.91, and the "Average Comparison" column shows the letter "B". In this case, it is significant at the significance level of 0.05 because it is shown in upper case. Regarding the degree of satisfaction with the seminar itself, there was a significant difference in the score between general and those with a position.
Useful for between-group comparisons for many questions
In "Categorical", aggregation and comparison by group can be performed by specifying a grouping variable. In actual surveys, there may be many questions and many answer options. In such a case, many tables and graphs are displayed, so it is difficult to find out which answers have differences one by one.
If you use the "response homogeneity test" and "comparison of score averages" explained in this article, items with differences can be seen at a glance, so you can efficiently compare responses between groups.
In the next Part 3, we will explain in detail about "multiple response", which deals with multiple-response questions in questionnaires.
