Subscribe Bookmark RSS Feed

Two-way ANOVA 'error'

lholli8

Community Trekker

Joined:

Jun 9, 2016

I am trying to run a two-way ANOVA (unbalanced) with two main effects: Effect A has 7 levels (6 + control); Effect B has 4 levels (3+control). When I run a full factorial, JMP 'ignores' the controls or I get that 'Lost DFs' in the ANOVA table and no results. What is going ON? It has something to do with the controls...

5 REPLIES
txnelson

Super User

Joined:

Jun 22, 2012

Not seeing the data table you are using makes it pretty tough to diagnose the issue.  Can you attach a copy of the data table.  If the data responses are confidential, you can make the response data random values.  What is critical are the values of Effect A and Effect B.

Jim
lholli8

Community Trekker

Joined:

Jun 9, 2016

11803_JMP data example.png

Hi Jim,

Thanks for the reply and sorry for the delay! I had to leave out column headings for 'security' reasons. Column 1 is the sample number (e.g. 76-9). Column 2 is Effect A (e.g. HN) and Column 3 is Effect B (e.g. Control). Effect A has 7 levels, Effect B has 4 levels. There are several response variables in the table. During the initial two-way ANOVA run, I used Column 11 (fourth from the right, e.g. 1.2 at the top). The problem seems to rest with the control level. By the way, there are 10 observations (or tested subsamples) of each replicate; for instance, there are 10 HN x Low x 4 replicates = 40. Ditto with Control.

I hope this makes sense. This is an interactive study where both treatments (A and B) were added to each exp. unit. Thanks in advance. -LH

markbailey

Staff

Joined:

Jun 23, 2011

I interpret 'run a full factorial' to mean that your model has terms for the two main effects and the interaction effect. I interpret 'ignored controls' to mean that you do not see estimates for this level of your factor. The parameterization of the model for categorical factors in JMP is such that the last level (think JMP value ordering) is not reported because it is equal to the negative of the sum of the estimates of the other levels. That is, all of the estimates must sum to zero.

Click the red triangle at the top next to Fit Least Squares and select Estimates > Expanded Estimates to see the results for the last level.

Learn it once, use it forever!
lholli8

Community Trekker

Joined:

Jun 9, 2016

Hi Mark,

Thanks for the reply and sorry for the delay! (This is the same message to Jim, above) I had to leave out column headings for 'security' reasons. Column 1 is the sample number (e.g. 76-9). Column 2 is Effect A (e.g. HN) and Column 3 is Effect B (e.g. Control). Effect A has 7 levels, Effect B has 4 levels. There are several response variables in the table. During the initial two-way ANOVA run, I used Column 11 (fourth from the right, e.g. 1.2 at the top). The problem seems to rest with the control level. By the way, there are 10 observations (or tested subsamples) of each replicate; for instance, there are 10 HN x Low x 4 replicates = 40. Ditto with Control.

I hope this makes sense. This is an interactive study where both treatments (A and B) were added to each exp. unit.

11804_JMP data example.png

Thanks in advance! -LH

lholli8

Community Trekker

Joined:

Jun 9, 2016

Mark,

I can paste this table because it doesn't reveal what I'm studying:

11805_Interaction Least Squares 2-way ANOVA.png

By the way, 'C' is also 'Control' - I thought that it would help if I changed one of them. This is the interaction table - similar results for Effects A and B, i.e. Control interactions are 'NonEstimable'. Thanks. -LH