How to get a single output table for one-way ANOVA on many response variables

[#1]

Hello,

I'm trialing JMP and a few other stats software packages before committing. As an environmental scientist, I routinely need to run 100 one-way ANOVAs to test for differences in a suite of ~20 water chemistry analytes between yearly sampling events (n = 4) for five test locations (6 replicates within each test area). I'm not testing for differences between area, only differences in mean concentrations of each analyte between each of the 4 events (yearly). I need the one-way ANOVA result for year difference for each of the 20 analytes, with the dataset split by area. I have found how to make the JMP output give me the area split (good), but within each area split it has 20 drop downs (i.e., one for each analyte) which I then need to go through individually (so, 100 times total), to select the ANOVA result from the dropdown menu. This is not so fun. Is there any way to make this less tedious? How can I get all 20 ANOVA results for each area as a single table?

Cheers

[#2]

There is also a "fun" version to do this

with 

20 analyted, 5 areas  -> :age

sampling event -> :sex

just run the script and then right click on one of the Anovas and select Make combined Data Table

Names Default to Here(1);
dt = Open( "$SAMPLE_DATA/Big Class.jmp" );
gb = dt << Graph Builder(
	Size( 437, 413 ),
	Graph Spacing( 4 ),
	Variables( X( :height ), Y( :weight ), Overlay( :sex ) ),
	Elements( Points( X, Y, Legend( 1 ) ), Smoother( X, Y, Legend( 2 ) ) )
);
	
Oneway(
	X( :sex ),
	Y( :height ),
	By( :age ),
	SendToByGroup( Bygroup, Y( :height ) ),
	SendToByGroup( Bygroup, X( :sex ), Means( 1 ), Mean Diamonds( 1 ) ),

);

[#3]

I believe the easiest way to handle this is to use the Response Screening Platform

     Analyze=>Screening=>Response Screening

It gives you the One Way ANOVAs in a table format

[#4]

I wonder about another approach altogether. Each Analyte (20) has a unique response (e.g., dissolved oxygen), range, and variance, so let's keep them separate for now. You can enter all of them in the Y role with Fit Model for parallel analysis and reduce your effort. You could build a linear predictor with Location (5 levels) and Year (4 levels) to further reduce your effort and obtain more information. I am not sure yet about how to use the replicates (implicit or explicit random effect term). Why do it this way?

• More efficient for you. You specify one platform launch and get all the results at once.
• More efficient for JMP. You lose statistical power and accuracy by dividing the data into many smaller sets.
• You can ask more questions of the data.
  • Is the effect of time the same at each location?
  • Do some locations behave the same? Are there clusters that might inform you?
• You can see all the statistical tests together.
• You can perform post hoc custom tests or contrasts.
  • You can ask questions about individual locations or years, or different combinations of either variable.

How do you set it up for this kind of analysis? You must have a data column for each Analyte response and a data column for the Location and Year. Then follow these steps:

1. Select Analyze > Fit Model.
2. Select the Analyte data columns and click Y.
3. Select the Location and Year data columns, and click Add.
4. Set the Emphasis to Effect Leverage.
5. Click Run.
6. Examine the Analysis of Variance table to determine if the whole model is significant compared to the grand mean.
7. Examine the Effect Tests to determine if the Location or Year is significant.
8. Click the red triangle next to the Leverage Plot for Location and select LSMeans Plot and LSMeans Tukey HSD.
9. Repeat the last step with Year.

This analysis should be a good start. I hope you find it helpful.