Cohen's d

[#1]

Hello

I conducted one-way ANOVA to compare the average of a variable across 3 groups. The reviewer of my paper asked for Cohen's d test for measuring the effect size. How can I get this measure in JMP? I am using JMP pro V17

[#2 · Accepted solution]

Cohen's d is not directly available in JMP 17.  However, it is in JMP 18 under

     Analyze=>Fit Y by X=>Means/Anova/Pooled t

In JMP 17 it has to be calculated using JSL.  Here is an example script that generates Cohen's d for all pairs of groups

or

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

var = "height";
grp = "sex";
Eval(
	Parse(
		"Summarize(
		dt,
		bygp = by( :" || grp || " ),
		stds = Std Dev( :" || var
		 || " ),
		ns = Count( :" || var || " ),
		means = Mean( :" || var || " )
	);"
	)
);

nw = New Window( "Cohen's d", vlb = V List Box() );
For( i = 1, i <= N Items( bygp ) - 1, i++,
	For( k = i + 1, k <= N Items( bygp ), k++,
		pooled = Sqrt(
			(stds[i] ^ 2 * (ns[i] - 1) + stds[k] ^ 2 * (ns[k] - 1)) / (ns[i] + ns[k] - 2)
		);
		dif = Abs( means[i] - means[k] );
		cohen = (dif / pooled);
		stats = Matrix( dif ) |/ Matrix( pooled ) |/ Matrix( cohen );
		compare = {};
		insert into( compare, bygp[i] || " vs. " || bygp[k]);
		ob = Outline Box( grp || "=" || bygp[i] || " vs. " || grp || "=" || bygp[k],
			Table Box(
				String Col Box( "", compare ),
				String Col Box( "", {"Difference", "Pooled Std Dev", "Cohen's d"} ),
				Number Col Box( "", stats )
			)
		);
		vlb << append( ob );
	)
);

[#3]

Thanks 

It was helpful. I wonder if there is a way to get it in JMP 18 for 3 groups, not 2. pooled t is only available when comparing means in 2 groups, however, in my case, I conducted one-way ANOVA to compare means of 3 groups

[#4]

Hi @Samira,

Outside of 2 groups, this option is not possible with JMP 18 natively. You can have a look at the Wish list, this need has already been raised and acknowledged : Cohen's d for "Fit Y by X" 

Let's hope for a release in a future 18.X version or JMP 19.

[#5 · Accepted solution]

Hi @Samira : FWIW, for 2 or more groups in the ANOVA, the RMSE (Root Mean Squared Error) is the pooled SD. 

[#6]

Thanks

This helped me a lot to calculate Cohen's d manually. For more clarification, I found RMSE in the summary of fit table.