topic Display regression estimates' F-ratios? in Discussions

Display regression estimates' F-ratios?

MAS — Sat, 16 May 2020 17:45:15 GMT

Is there an option to display the F-ratio (square of the t-ratios) in the regression display of the *individual* estimates in Fit Least Squares? I know JMP displays the overall model F-stat, but I'm looking for the F-ratios for each independent variable alone.

Re: Display regression estimates' F-ratios?

txnelson — Sat, 16 May 2020 18:41:17 GMT

In the Parameter estimates output from the Bivariate Platform, there is not a column that is available to be displayed that will show the F Ratio. If you are using a different platform you can right click on the table box in question, select Columns and see if an F Ratio column is available to be displayed.

Regardless, you can add your own F Ratio column to the table box. Below is a simple example of how to do that

Here is the simple script that added the F Ratio

Names Default To Here( 1 );
dt = Open( "$SAMPLE_DATA/big class.jmp" );

biv = dt << Bivariate(
	Y( :height ),
	X( :weight ),
	Fit Line( {Line Color( "Medium Dark Red" )} )
);

// Get the t ratios
theTs = Report( biv )["Parameter Estimates"][Number Col Box( 3 )] << get;

// Convert them to F Ratio values
theTs[1] = theTs[1] ^ 2;
theTs[2] = theTs[2] ^ 2;

// Append them to the output Table Box
theTs = Report( biv )["Parameter Estimates"][Table Box( 1 )] <<
append( Number Col Box( "F Ratio", theTs ) );

Re: Display regression estimates' F-ratios?

statman — Sat, 16 May 2020 22:57:57 GMT

I may be a bit confused, but the F-ratios are displayed in Analysis of Variance (which provides the whole model statistics) and in the Effects Tests outputs of the Fit Model platform. It sounds like you want the Effects Tests which decompose the model DF's into each DF.

Re: Display regression estimates' F-ratios?

ron_horne — Sat, 16 May 2020 23:01:52 GMT

Hi @MAS ,

if you are using the fit model platform with a continuous dependent variable you can get make use of the Effect tests table. This gives an F tests for each effect. for continuous variables this is identical to the t test. the same for nominal or ordinal variables with only two categories. form more than two category variable this F tests is a bit different than the individual parameter estimates.