When we interact with JMP users on a regular basis, we sometimes come across features that we would never have thought of as "good features of JMP."
The function for specifying the "By" variable in JMP that I will introduce this time is a function that I use on a daily basis, so I have the preconceived notion that it is something that is obvious and so have not particularly highlighted it until now.
However, in recent years, we have received comments from several JMP users saying that the By function is "very useful" and "very helpful when using JMP." In this article, we will explain examples of how to use the By function and the useful functions that have been newly added to the latest version, JMP 18.
Finally, we will also introduce a user case study that was recently published in Japan that demonstrates the usefulness of By variables.
By変数の指定とそのレポートSpecifying By variables and reporting them
Specifying By variables and reporting them
In JMP, many platforms in the Analyze menu allow you to select a column as the By column to perform analysis by category of a specified variable, making it easy to perform what is known as "stratified analysis."
For example, let's say you want to use the JMP sample data "Powder Metallurgy.jmp" to see if there is a difference in "stretch rate" depending on the "molding method." If you want to look at each type of powder (5 types), you would specify the column "Powder type" as [By] in [Relationship between two variables].
The report is split into powder type categories (Type 1 to Type 5).
You can also specify multiple columns in [By]. If you specify multiple columns, a report will be displayed for each combination of the categories of those variables.
Useful functions when using By variables
In JMP 18, a new feature called "Group Platform" has been added. When you specify a By variable or multiple Y variables, many reports are displayed at once, but with this feature, you can analyze them together.
The figure below shows a comparison of reports when specifying a By variable in "JMP 18" and "JMP 17 and earlier." In JMP 18, an outline called "One-way analysis of stretch rate by molding method" is created, which contains reports Type 1 to Type 5. You can apply analysis options to all reports at once by clicking the red triangle button in the upper left corner of this outline.
In JMP 17 and earlier, if you wanted to apply an option to all reports, you had to Ctrl+click the red triangle and select the option. Also, when you applied a local data filter or column switcher to a report, you could not globally apply it to all reports.
New Features in JMP 18
Here are three useful features when using By variables in JMP 18:
1. Change the layout
Using the By variable will display reports one above the other by category, but if you have a large number of categories, scrolling can become tedious.
In this case, you can use the Layout option to arrange them in multiple rows or display them in tab format, as shown below.
2. Apply options to all reports at once
As mentioned above, you can apply options to all reports at once by clicking the red triangle in the upper left corner of the report.
Below is an example of the Means/ANOVA option applied to all reports.
Local data filters (a function for filtering reports) and column switchers (a function for switching the columns to be analyzed) can also be applied in bulk.
3. Sort the report by relevance
The Order option allows you to order multiple reports by the goodness of fit (R-squared). The following report shows what happens when you use the Order > Goodness of Fit, Descending option. The reports are sorted from left to right by the largest R-squared.
If the number of categories in the By variable is large, the number of reports displayed will be large. In that case, it is useful to use this function to sort by the best fit.
User Case Studies
Recently, we published the case study of Energywiz Co., Ltd. as a JMP user case study.
In this example, the usefulness of the By variable is stated as follows:
"You can quickly create a histogram using the 'Univariate Distribution' and then use the 'By Variable' to draw histograms for each month or product at once. I often start by checking where a certain variable's population is located on the histogram."
I feel that this is a very useful example of how to use JMP.
Note
Using by variables to dig deeper into the data is useful from the perspective of exploratory data analysis. However, because many associations are examined at once, care must be taken when conducting tests to avoid issues such as multiplexing and reduced power due to a reduction in sample size.
by Naohiro Masukawa (JMP Japan)
Naohiro Masukawa - JMP User Community
