作業効率アップ！「モデルのあてはめ」のレポートカスタマイズ術

[#1]

The JMP Fit Model platform lets you build a variety of models, including regression, analysis of variance, logistic regression (nominal and ordinal), and generalized linear models.

The Fit Model platform is used to fit regression models to observed data or to fit response surface models to experimental data designed using Design of Experiments methods.

When you specify a continuous model for Y in the Model Specification window, the default method is Standard Least Squares. But do you know how this determines which graphs and tables are initially displayed in the Report window?

How emphasis determines which report is displayed first

In fact, the items specified in "Emphasis" determine which table or graph is displayed first.

Below are the graphs and tables that will be displayed initially for each "Emphasis" item.

The default "emphasis" is determined by the number of rows in the data (n) and the number of terms included in the model effects (k), according to the following rules:

• If n ≤ 1000, Emphasis is set to Minimum Report.
• If n ≤ 1000 and k ≤ 4, Emphasis is set to Effect Leverage.
• If n ≤ 1000 and k ≥ 10, Emphasis is set to Factor Screening.
• If n ≦ 1000 and 4 < k < 10 and n - k ≧ 20, Emphasis is set to Effect Leverage.
• If none of these conditions are met, then Emphasis is set to Screening Factors.

Note : This does not include rules for when a random effects model is specified.

For example, when designing an experiment to screen for main effects and second-order interactions with five factors, if you use [Custom Design], the default number of experiments will be 20. In this case, the number of terms to be examined will be 15, including the main effects (5) and second-order interactions (10), so the third rule above will be applied and it will be set to [Screening of Factors].

It is convenient for JMP to automatically display the required report based on the number of data and items, but there are times when you want to display a specific table or graph.

For example, by default, Factorial Screening does not display the Summary of Fit or Analysis of Variance reports, and some of these statistics are displayed in the Actual vs. Predicted plots, but you may want to see the Adjusted R-Square for model comparisons.

In this case, you will need to add the report by selecting the appropriate option from the red triangle button in the upper left corner of the report.

Additionally, Effect Leverage does not appear in the Prediction Profiler report by default, you must add it to the report by specifying the option as well.

If you use Fit Model frequently, specifying the options each time can be a little tedious, so why not take advantage of the ability to save your preferred reports as presets?

Save your favorite reports as presets

In the latest version of JMP 18, you can save the tables and graphs displayed in a report as a preset and apply it to other reports.

In the example below, I have included all the tables and graphs that I consider in "Fit Model". The confidence intervals (lower 95%, upper 95%) and VIF for the "Parameter Estimates" can be accessed from the right-click menu in the report.

Once you have added all the reports you want to display, click the ★ icon to the right of the report and select [Save Preset As] . You can give the file a name of your choice when saving.

Once you have saved a preset, you can apply it to other Fit Model reports.

Sharing presets

Saved presets can be saved as files and shared among JMP users.

From the menu [File] > [Preferences], select the "Platform Presets" category, select the preset you want to save, and then click the [Export] button.

Save the preset with a file name of your choice. The preset file extension will be "jmppresets".

You may want to share the saved preset file with JMP users on different machines or within your organization to ensure that the reports they view are consistent.

The shared file can be loaded using the [Load] button in the preferences mentioned above, and the presets can be applied on that machine.

For reference, I will share the presets for the "Fit Model" report that I saved on this blog.（fitmodel_presets.jmppresets）。

You can use these presets as a base to create your own presets.

If you frequently use JMP's Fit Model feature, you'll find this preset to be a great way to speed up your work.

Note : In JMP 18.0.x, there is a problem that prevents you from saving a model fit profile as a preset. To apply a preset, use JMP 18.1.x or later.

reference:

New Features in JMP 18 Part 3: The popular "Platform Presets" feature - JMP User Community

by Naohiro Masukawa (JMP Japan)

Naohiro Masukawa - JMP User Community