topic Re: Desperate help on non-normally distributed tests in JMP! in Discussions

Desperate help on non-normally distributed tests in JMP!

PedCards2017 — Fri, 09 Jun 2023 01:01:21 GMT

I am fairly junior in the stats realm and am trying to figure most of this out on my own.

My data is non-normally distributed (already ran the Shapiro Wilke normality test) and I have multiple variables I am trying to compare between 2 groups over multiple time periods with p-valves for each time period and each variable. Some of the reading I have done seems like I should run a Wilcoxon rank sum test but having trouble setting it up correctly in JMP (even with the help section online) or if anyone else knows if I should be using another type of test I would appreciate it!

Re: Desperate help on non-normally distributed tests in JMP!

Mark_Bailey — Wed, 26 Apr 2023 20:04:34 GMT

When you say, "I have multiple variables I am trying to compare between 2 groups over multiple time periods," are you trying to see if any variable can distinguish between the groups? Also, are the time periods the same for all variables across both groups?

Re: Desperate help on non-normally distributed tests in JMP!

MRB3855 — Thu, 27 Apr 2023 07:53:32 GMT

You say "My data is non-normally distributed"; it is not the "data" that should be normally distributed; it is the residuals that should be normally distributed. Assuming your model is correct (sounds like it may be a repeated measures design), have you looked at the residuals? As @Mark_Bailey said, can you elaborate on your design, data, and goals?

Re: Desperate help on non-normally distributed tests in JMP!

PedCards2017 — Fri, 28 Apr 2023 00:14:48 GMT

I am trying to look to see if there is any difference in the variables between the 2 groups with multiple different variables but not distinguish the group. Just trying to see if there is a significant difference among the variables between the 2 groups. And the time periods are the same across all the variables.

Re: Desperate help on non-normally distributed tests in JMP!

PedCards2017 — Fri, 28 Apr 2023 00:16:10 GMT

See above, but I am doing a retrospective cohort study looking at group A and B and multiple measurements of echocardiography to see if I can detect differences in these measurements between the 2 groups over different time periods.

Re: Desperate help on non-normally distributed tests in JMP!

MRB3855 — Sat, 29 Apr 2023 06:24:18 GMT

Ok. So, to clarify. You have two groups, A and B. Group A has n subjects, and group B has m subjects (m may or may not equal n). And each subject has multiple measurements of echocardiography (I.e., for each subject, ecg is measured at different times say t1, t2, …, tp). Do I have this correct?

Re: Desperate help on non-normally distributed tests in JMP!

peng_liu — Sun, 30 Apr 2023 18:23:35 GMT

I am going to use Big Class as an example to illustrate one approach.

If you don't know what "Big Class" is, first go to JMP menu and click "Help", then "Sample Index", find the button in the following screenshot and click it. It will open "Big Class" data table.

Now go to JMP's Analyze menu, find "Reliability and Survival" sub-menu, then select "Life Distribution". Configure the dialog as follows:

This is to illustrate how to compare "weight" between two genders. Here is the report.

There can be different perspectives, depending how you would like to compare them, and what are important to you. From the perspective of whether the distributions of "weight" are significantly different between male and female, the result says nay.

Re: Desperate help on non-normally distributed tests in JMP!

PedCards2017 — Sun, 30 Apr 2023 20:34:01 GMT

Yes, that is correct.

Re: Desperate help on non-normally distributed tests in JMP!

PedCards2017 — Sun, 30 Apr 2023 20:36:13 GMT

I am willing to try. I initially was trying to get whisker-box plots with p-values between each "box" between the 2 groups as a way to display the data. But I can try this as well!

Re: Desperate help on non-normally distributed tests in JMP!

MRB3855 — Sun, 30 Apr 2023 21:59:48 GMT

This is a classic repeated measures design. There is a lot to consider, but may I suggest you start here.

https://www.jmp.com/support/notes/30/584.html

Re: Desperate help on non-normally distributed tests in JMP!

PedCards2017 — Mon, 08 May 2023 13:58:52 GMT

That didn't really work for my data. Also, any reason why a column of numerical data is showing up as ordinal in the tables when I go to analyze? Do i need to enter something into blank boxes in order to not analyze those?

Re: Desperate help on non-normally distributed tests in JMP!

Mark_Bailey — Mon, 08 May 2023 14:37:25 GMT

In what way "didn't (it) really work?"

Did the numerical data begin with the character data type? Did you convert it to the numeric data type? If so, you must also switch the modeling type from nominal to continuous.

Re: Desperate help on non-normally distributed tests in JMP!

MRB3855 — Mon, 08 May 2023 14:44:12 GMT

What do you mean by “that didn’t really work for my data”. What is the “that” that didn’t work, and what do you mean by “didn’t really work”.

As for the ordinal issue, you’ll have to make sure the data are in the correct format (https://www.jmp.com/support/help/en/17.0/index.shtml#page/jmp/column-properties-in-jmp.shtml).

Therein it says:

To assign a column property to one or more columns

1.Select the column or columns to which you want to assign a property.

2.Do one of the following:

–Right-click the header area, select Column Properties, and select the property.

–Right-click the header area, select Column Info, and select the property from the Column Properties menu.

–Select Cols > Column Info and select the property from the Column Properties menu.

3.In the column property panel that appears, specify values and select options as appropriate.