Fit Model and Custom Estimates

[#1]

Hi!

I am using JMP 16 and trying to use Fit Model -> Custom Estimates and I followed this JMP help page https://www.jmp.com/support/help/en/16.2/#page/jmp/custom-test.shtml#ww1531879

However, I don't understand why one of the comparisons is not shown, the example aims at comparing:

"Suppose that you want to test three contrasts. You want to compare the mean responses for the following:

•treatment A to treatment B

•treatments A and B to the control group

•treatments A and B to the control and placebo groups"

But in the output I only see 3 comparisons, the "Placebo" group is not there. And when I tried to run this or my data set which also includes 4 treatments, it also happens that one of the treatments does not appear in the editable table for the custom test.

Can anyone please help me with this?

Thanks.

[#2]

The example of three custom tests (contrasts) does not include the Placebo level in any contrast, so there are only three tests and there is no comparison with that level. You have to specify exactly which comparisons you want, including the Placebo level. The level 'Placebo' has not meaning to JMP. It is just a label for a unique factor level. It is not automatically considered in any custom test.

What are the comparisons that you want to make?

[#3]

Hi! 

I wondered about the Placebo on the example because that's one of the options to be tested, as I pasted above and copied from the JMP help page.

Anyways, I have 3 treatments on my data set. After running the fit Model and when I get the editable table for the Custom test, only 2 of the treatments appear as rows in the table. I wonder why I don't see the 3rd treatment because then I can't include it in the comparisons.

[#4]

Can you show the results of your tests? Did you get an error message? (You might check the Log.)

There are some joint custom tests that are inestimable because they introduce a constraint on the linear contrast. That is, the tests must be independent to be estimable.

[#5]

No, I didn't get a error message.

So I have 3 treatments: Control, Fire Severity and Salvage-Logged.

I would like to compare "Control vs Fire Severity", "Control vs Salvage-logged", "Fire Severity vs Salvage-logged" but since the "Salvage-logged" treatment does not appear in the table I can't run this comparison. This is what I don't understand.

Thanks for your help!

 

[#6 · Accepted solution]

Click the red triangle next to the Treatment leverage plot and select Turkey HSD. This command will perform all pair-wise comparisons with an adjustment for the number of comparisons.

[#7]

Alternatively, you might use Analyze > Fit Y by X > Oneway for this simple case and then select Compare Means > Tukey from the red triangle menu.

You do not really need a custom test in this case of all pair-wise comparisons.

[#8]

I am wondering why I can't do the test with all the treatments I have in the data set, since one of them doesn't show up in the editable table, as you see on the picture I added above. It would be of great help to get some insights on this. Thank you so much!