turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

- JMP User Community
- :
- Discussions
- :
- Definitive Screening Design Categorical Factor Lev...

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Sep 8, 2017 3:28 PM
(1352 views)

I am working on a Definitive screening Design (my first) I set design up with 7 factors one of which was a categorical (equipment type) with two levels. When reviewing design with team, request was made to skip step for this equipment two set one of the continous factors to 0.

I am reviewing design now and this change really makes the categorical variable into a 3 level item. I am concerns about issues this may cause with results analysis, and if there are any special steps I should take to modify the design to handle this item?

Any help or suggestions are appriceated.

Thanks, Chris

1 ACCEPTED SOLUTION

Accepted Solutions

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Sep 13, 2017 4:12 AM
(2380 views)

Solution

Okay.

I had a quick look at the consequence of adding the 3rd level to your categorical X1 when X2 = 0. It does have some significant effects on the properties of your design, as you might expect. I would encourage you to have a look for yourself using the Design Evaluation platform.

Also, I am not sure that it really helps with interpretation of the resulting model. Your resulting model would cover possibilities where X1 = L3 (i.e. no machine used) and X2 > 0, which doesn't make a lot of sense.

You could just keep the design as is and not worry that X1 is essentially meaningless when X2 = 0. The term for the interaction of X1 and X2 should be able to capture the fact that there is no difference between levels of X1 when X2=0.

Or you could reframe the problem or rethink factor ranges so that they make more sense.

Or you could go to a Custom Design.

I hope this helps.

Cheers,

Phil

I had a quick look at the consequence of adding the 3rd level to your categorical X1 when X2 = 0. It does have some significant effects on the properties of your design, as you might expect. I would encourage you to have a look for yourself using the Design Evaluation platform.

Also, I am not sure that it really helps with interpretation of the resulting model. Your resulting model would cover possibilities where X1 = L3 (i.e. no machine used) and X2 > 0, which doesn't make a lot of sense.

You could just keep the design as is and not worry that X1 is essentially meaningless when X2 = 0. The term for the interaction of X1 and X2 should be able to capture the fact that there is no difference between levels of X1 when X2=0.

Or you could reframe the problem or rethink factor ranges so that they make more sense.

Or you could go to a Custom Design.

I hope this helps.

Cheers,

Phil

8 REPLIES

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Sep 12, 2017 3:53 AM
(1319 views)

Hi Chris,

I am not quite sure what you mean. Could you send the design. If necessary you could create an anonymised version jsut with X1... X7 and coded factor levels (-1,0,1).

Regards,

Phil

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Sep 12, 2017 6:51 AM
(1307 views)

Please see attached design example.

[cid:image001.png@01D32B9C.310881D0]

[cid:image001.png@01D32B9C.310881D0]

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Sep 12, 2017 8:14 AM
(1293 views)

Unfortunately, images in email replies don't make it to the Community. @chris_dennis, could you edit your response above to attach the image?

-Jeff

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Sep 12, 2017 8:39 AM
(1287 views)

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Sep 12, 2017 8:56 AM
(1278 views)

That is interesting. So you have a categorical X1 (L1, L2) and a continuous X2 (0, 8000, 16000). And if X2 is 0, then X1 is neither L1 nor L2.

Does it make sense to define X2 as a continous over the range 0 to 16000? After you have run the experiment and modelled the data, will it make sense for the model to make predictions of the response for X2 = 1 or X2 = 500? Would you actually operate in the range 1 to 8000?

Could you change X2 to have a range of 1 to 16000? Or some other range with a lowest setting > 0.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Sep 12, 2017 10:06 AM
(1273 views)

It is possible to do a low value like 1000, but not easy to process.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Sep 13, 2017 4:12 AM
(2381 views)

I had a quick look at the consequence of adding the 3rd level to your categorical X1 when X2 = 0. It does have some significant effects on the properties of your design, as you might expect. I would encourage you to have a look for yourself using the Design Evaluation platform.

Also, I am not sure that it really helps with interpretation of the resulting model. Your resulting model would cover possibilities where X1 = L3 (i.e. no machine used) and X2 > 0, which doesn't make a lot of sense.

You could just keep the design as is and not worry that X1 is essentially meaningless when X2 = 0. The term for the interaction of X1 and X2 should be able to capture the fact that there is no difference between levels of X1 when X2=0.

Or you could reframe the problem or rethink factor ranges so that they make more sense.

Or you could go to a Custom Design.

I hope this helps.

Cheers,

Phil

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Sep 18, 2017 12:27 PM
(1142 views)

I change to the Customer Design platform. Because of the interaction between X1 and X2 I combined these factors into 5 descrete levels. Thanks for the help.

Chris