in the previous blog A 2-minute video introducing the concept of "categorical" (Japanese) was introduced, and many viewers commented that they were interested in the categorical function and that they would like a longer video, so please explain the details in detail.
Therefore, in this blog series (3 parts in total), we will talk more about the characteristics of "categorical" based on analysis examples of questionnaires.
Although it looks complicated, it is very convenient for analyzing questionnaires.
From the JMP menu bar Edit > Consumer Research > Categorical Selecting launches a dialog to select the role of the response, as shown below. At first glance, it is confusing because there are many tabs and buttons.
Just don't give up here! If you know only the basics of how to use this "categorical", you can use it very effectively in analyzing surveys.
Classification of multiple choice answers
Questionnaire answers are mainly divided into single answers (only one answer from the choices) and multiple answers (you can choose multiple answers from the choices). In addition, considering the JMP scales (nominal, order, continuous), there are nominal scales that respond to gender and residential area in single response, and ordinal scales that respond to age and product usage frequency, etc. One example is a continuous scale that scores products out of 10.
This time, we will use the questionnaire answered by the participants in the in-house SQC (statistical quality control) training as an example, but it can be classified as follows as single answers and multiple answers.
Below is a summary of the survey results in a JMP data table. In this example, the multiple answers "Q1 Why I took the course" (yellow column) are entered in a single cell separated by commas. For example, if a respondent answered "necessary for work" and "further understanding of SQC", the respondent's The cell contains "Required for business, better understanding of SQC".
Specify "response role"
"Categorical" specifies how you want the columns in the data table to be treated. "Response role" will be specified as Various settings can be made, but when analyzing survey data, it is good to remember the following three (simple response, response with multiple values, and multiple response).
simple: response
Used for single answer. In this example, we have specified the respondent's attributes (position) and level of understanding before the training.
Related: responses with common values
A survey may ask several questions with the same answer choices. In this example, the level of understanding and satisfaction with the outline of the training, the level of understanding and satisfaction with the basics of the training, and so on are evaluated on a 5-point scale. You might want to compare these questions, such as high level of general understanding but low level of satisfaction with basics. In such a case, from the "Related" tab [response with common value] Select to display these multiple responses in a single table or graph.
Multiple response: delimiter
The Multiples tab is useful when aggregating multiple responses. There are various data input formats for multiple answers, and it is characterized by being able to handle any format without processing the data. (More on this in Part 3.)
In this example, the data is entered in one cell separated by commas (,). [multiple response delimiter] Choose.
"Categorical" reports
"Categorical" reports display summary tables and share charts (corresponding to bar graphs that show the proportion of responses within each group).
Note the following distinctions for multiple responses in the tabulation:
Total number of responses: total number of responses (= 674)
Total number of cases: number of respondents (= 503)
The total number of responses is the total number of responses given by one respondent. If everyone answered all three options, the total number of responses would be three times the number of respondents.
Each cell in the summary table consists of three rows, the second row showing shares (frequency ÷ total number of responses) and the third row showing cases (frequency ÷ total number of cases). Share is where all categories add up to 100%.
This is the report for the responses specified in "Responses with Common Values". Both the summary table and share chart are combined into one table and graph, making it easier to compare questions.
Also, if the questions are scored, it may be easier to understand if the average score is used to compare each question. For Categorical, select Average Score or Standard Deviation of Score from the red triangle at the top left of the report to display the mean and standard deviation on the right side of the table.
Considering the average scores, we can see that comprehension and satisfaction with the exercises are higher for the other items.
Statistical analysis beyond aggregation
What we have introduced so far is a method for aggregating individual questions. In questionnaire analysis, it is common to compare answer options by dividing them by attributes. In the next Part 2, I will introduce how to divide into groups by "categorical" and compare them statistically.
by Naohiro Masukawa, JMP Japan
