cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
Choose Language Hide Translation Bar
Sandeep123
Level III

Calculating effect size

I have a data set of about 2000 patients. There are two groups one with the critical illness and others with known critical illness.

I have done a univariate comparison between the two groups based on a there are age (continuous variable) and a number of other categorical variables like sex or race comorbidities et cetera.

Because of the relatively large dataset most of the differences come back as statistically significant with the values less than 0.01.

The reviewers of my paper however want me to report and interpret standardised effect sizes (Cohan's D or Cramer's V).

It seems to be a pretty simple thing to do it on SPSS.

Cramér’s V - Beginners Tutorial (spss-tutorials.com)

 

It is possible to do this in JMP? The only documentation I could find online was for doing it in a standard least square model in linear regression

11 REPLIES 11
ron_horne
Super User (Alumni)

Re: Calculating effect size

Hi @Sandeep123 ,

perhaps this discussion help you calculate it manually if you just need a few estimates.

https://community.jmp.com/t5/Discussions/How-to-get-a-p-value-on-descriptive-statistics-How-to-get-C...

 

Ron

 

ron_horne
Super User (Alumni)

Re: Calculating effect size

Hi @Sandeep123 ,

perhaps this discussion help you calculate it manually if you just need a few estimates.

https://community.jmp.com/t5/Discussions/How-to-get-a-p-value-on-descriptive-statistics-How-to-get-C...

 

Ron

 

Sandeep123
Level III

Re: Calculating effect size

Ahhh

 

so there is no way to get this automated in jmp. Some other marker of effect size may be?

Thierry_S
Super User

Re: Calculating effect size

Hi Sandeep,
It looks like there might have been some confusion about your question. The Cohen's D effect size measure is simply the difference between the means of the group you compared divided by the pooled standard deviations. While this measure is not directly available in JMP (at least in version 14.1), there is a JMP Add-in that does exactly what you need: Calculate Effect Sizes Add-in .
I'll continue to dig to see if there is something even simpler to address your needs.
Best,
TS

Thierry R. Sornasse
Sandeep123
Level III

Re: Calculating effect size

yes it is only available as LSM model not in Fit X by Y platform..which is fine. I can use it for cohen's D. Cohen's D is also very simple to calculate by hand.

 

Problem is with categorical variables..Cramers' V ?

Sandeep123
Level III

Re: Calculating effect size

It seems to me that the measures of association option available in JMP is what I need

 

It does not give Cramers V but gives Lambda and un certainty coefficient..which seems to be similar

 

Measures of Association (jmp.com)

 

Crosstabs statistics (ibm.com)

Re: Calculating effect size

The two measures that you asked about are not available in JMP, but many others are available. See the Help entry. Perhaps one of them might satisfy the reviewer.

Sandeep123
Level III

Re: Calculating effect size

Sandeep123_0-1617043039276.png

yes .JMP give these measures for association.

 

Seems from the help menu that lambda and uncertainty coefficient best suit for nominal data. 

 

The values however are very low for lambda it is mostly 0.0000 even when p value is < 0.001

 

Since i have not seen reported in literature, is it expected for them to be this low in retrospective clinical research

 

Re: Calculating effect size

The p-value may be used to assess the strength of the evidence against the null hypothesis (no association or measure is zero). The measure of association is used to assess the strength of the association (similar to measures of correlation). So you have strong evidence that the measure is not zero, but it is a weak association. There is a lot of uncertainty in the response even when you know a given category.

 

You know how to calculate V yourself from the reply above. What value of V do you get for the same association that you analyzed in the picture above?