nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

14782_en-US

liliconcom

A Question Using the Custom Design Disallowed Combinations Filter

I have a customer who wants to set up a design using the Custom Design Disallowed Combinations filter. 5 factors, all categorical, Yes/No for levels. They want to disallow both Yes/Yes AND No/No for two of the factors for any one treatment combination. In other words, for any one treatment combination, in the two factors, they must be at Yes/No...doesn't matter which is 'Yes' or which is 'No'...I tried cajoling the Disallowed Combinations filter with the following:

In other words disallowing Yes/Yes OR No/No and every time I try to Make the Design I get a 'failed to converge'. Any ideas on how to make this design?

Re: A Question Using the Custom Design Disallowed Combinations Filter

Thanks to all...why didn't I think of the estimability issue? Duh...

Re: A Question Using the Custom Design Disallowed Combinations Filter

I suspect you have interactions in the model. Take them out, they're not estimable.

Re: A Question Using the Custom Design Disallowed Combinations Filter

You might try combining the factors. Since they can only be yes/no or no/yes. They is a 100% correlation between them.

Pete

Sent from my iPhone

Re: A Question Using the Custom Design Disallowed Combinations Filter

Hi Peter,

I think the reason it fails to converge is because those 2 factors will then become completely confounded and therefore inestimable. You could just include 1 of these factors and then create the other factor using an IF formula. They still would not be estimable though.