How do I use a defined subset of items in my covariate list.

[#1]

The title is pretty much it. I have a list of 5 covariates, and 703 items on the list. Most are expensive so i want to choose only from say items 23-39 and items 267, but i want to bring the list along in the code in-case we have to bite the bullet and augment to more expensive items later on.

Is this doable?

Thanks

[#2]

I'm not sure I understand  your question.  What are you trying to do with the "covariates"?

Here are my initial thoughts:

Using several different regression type platforms, you could look for relationships between many variables.  Of course the concern with covariates is:

1. Measurement errors in the covariate,

2. Lag effect,

3. Frequency of changes in the covariate,

4. Multivariate concerns

5. Others...

I would start with reasonable hypotheses to narrow the field (critical thinking), then there are lots of options....depending on if you have a model in mind vs. no idea. 

I'd start with simple correlations: Multivariate Methods>Multivariate (put them all in).  The scatter plots are most important.

[#3]

Thanks for the reply! So my question is less about analysis of the effect of covariates and more about the initial design. So because of cost reasons, i need to limit the initial set of items that jmp can pick from the list to design with. So in the custom doe platform, i click the new button from JMP 16 under the factor box for covariates, and pick the factors i want for covariates. However, i can't restrict the list JMP chooses from in the final design steps without going back to the table and excluding the rows that correspond to the items i don't want to use in this first pass. JMP gives me the option to select ones i want included in the design, but doesn't let me require that those items are the only ones chosen. Again, i could exclude the other items at the table level, but then in the new table generated for the new design, it doesn't carry the rest of the 690 items that i excluded into the new table, so i can't use them for future augmentations easily.

That may have been too much detail, but does that make sense?

Thanks

[#4]

Well that certainly more information than the first post.  But, I still don't understand your issues with JMP and I still don't understand your situation well enough to provide good advice. There are still many un-answered questions. Perhaps one question is what do you mean by table...what new table?

Here is my current understanding...You are trying the create an experiment.  You want to include covariates from a large list.  You have some ideas as to which covariates you want to include.  You will be limited to the number of covariates to include in any model.  The analysis of covariates in a model requires degrees of freedom.  When incorporated into the analysis of an experiment, you have a mixed model (i.e., fixed and random effects in the mode)l.  You need to understand how this is analyzed, so analysis is important to planning.  I'm not sure what you mean by "future augmentation"?  You mean re-analyze with different covariates?  

OK, back to JMP...You are saying you can't specify which covariates to include?  How are you specifying them in JMP?  When you click on the covariate option, you  can select whichever columns you want from your data table.  Are you saying this isn't working?

If the columns for the other variables are still in the table, you can always go back and change what is included in the model.  Don't exclude them from the table, just don't add hem to the covariate list.

[#5]

---

@statman wrote:

Well that certainly more information than the first post.  But, I still don't understand your issues with JMP and I still don't understand your situation well enough to provide good advice. There are still many un-answered questions. Perhaps one question is what do you mean by table...what new table?

---

So when you do a design, you have the "make table" button at the end, and a new table pops up.

---

@statman wrote:

Here is my current understanding...You are trying the create an experiment.  You want to include covariates from a large list.  You have some ideas as to which covariates you want to include.  You will be limited to the number of covariates to include in any model.  The analysis of covariates in a model requires degrees of freedom.  When incorporated into the analysis of an experiment, you have a mixed model (i.e., fixed and random effects in the mode)l.  You need to understand how this is analyzed, so analysis is important to planning.  I'm not sure what you mean by "future augmentation"?  You mean re-analyze with different covariates?  

---

So by future augmentation, I mean that the subset of items in the covariate list that meet our preferred cost limits falls in a subset of the covariate space, and if we find that the optimal conditions fall outside of where we experimented, we would have to augment the experiment using the augment design window under the DOE tab to include more of the items. However, I only found that I can restrict the initial design to the set of items that fall in our budget by "hid[ing] and exclud[ing]" the other rows. This means that the full list doesn't get incorporated into the resulting table that generates after I hit the "make table" button at the end of the custom design menu, and thus can't be used for future augmentations.

---

@statman wrote:

OK, back to JMP...You are saying you can't specify which covariates to include?  How are you specifying them in JMP?  When you click on the covariate option, you  can select whichever columns you want from your data table.  Are you saying this isn't working?

---

Yea so like I said in the previous paragraph, its not the columns but the rows that are the issue. I'd like to only have certain rows to choose from for this first design, but in case we need them later on, still have the software carry the rest of the list over to the new table that generates when I hit the "make table" button at the end of the custom design menu. But when you do this with the hide and exclude method I mentioned earlier, the other 690 rows don't carry over, as you can see in the row index column that gets generated with the new table, where it acts like there's just a table with the rows that I didn't hit hide and exclude on.

Does that clear things up? If not, I can try to figure out how to do the action recorder and upload that.

[#6]

A set statement may be used to construct subsets that only contain a portion of the items of another set. The format is typical.

[#7]

---

@MartinaWalsh wrote:

A set statement may be used to construct subsets that only contain a portion of the items of another set. The format is typical.

---

Is this using JSL? I'm not entirely sure what your statement means.