Correlations with ordinals

[#1]

Hi,

From correlations results in design evaluation I don’t understand how you estimate the correlation of

• an ordinal and a continues factor and
•  the correlation between two ordinals.

If I understand correctly, the correlation between continues is the absolute value of Pearson correlation but what is it when ordinals are involved?

I can see that in ordinals the available correlations are (nr of levels) - 1 but I’m not sure how to calculate these correlations.

As example the ordinal variable X3 is represented by X3 1 and X3 2, but no X3 3.  
However when the X3 is evaluated against the other ordinal X4 the correlations are for the whole X3 vs specific X4 levels (i.e. X3*X4 1 and X3*X4 2 - so all X4 levels)

Thanks for your time.

EV

[#2]

The correlation of the estimates uses the columns in the model matrix. You have to expand the the column for this factor in the design matrix to the columns in the model matrix. The number of columns in the model matrix is 1 less than the number of levels. That is because the estimate for the last level is equal to the negative sum of the estimates for the other levels.

See the documentation about the parameterization of the linear model in JMP.

[#3]

Hi, 

Thanks for your answer.
However, I'm not sure I understand how to do that. I understand that is not as simple as to get the desing matrix and check for the correlations. 

I need to have the model matrix which means that I have to set a model firstly. But then, what? 

How to find the model matrix on JMP? 
And how to expand the model matrix using my DoE?

And then on expanded matrix what I do to find the correaltion? 

Thanks for your time

[#4]

I do not know why you want to calculate the correlation of the estimates when JMP provides this at the design stage and the analysis stage for you. Here is an example from the Help > Scripting Guide > Objects that illustrates how to get the model matrix X:

Names Default To Here( 1 );
dt = Open( "$SAMPLE_DATA/Drug.jmp" );
obj = dt << Fit Model(
	Y( :y ),
	Effects( :Drug, :x ),
	Personality( Standard Least Squares ),
	Emphasis( Minimal Report ),
	Run
);
G = obj << Get X Matrix;
Show( G );

[#5]

Hi,
Indeed, JMP can return many things but they are useless or dangerous if we cannot understand what they show. On the correlations output from design evaluation I was thinking that I see Pearson's correlations but just yesterday I realised that I was probably wrong. So I wish to understand what are these figures.

On the example you sent me I got the model matrix but I'm not sure how's this useful. I tried to check the correlation on the model matrix but doesn't seem to much the results from design evaluation.

On your 1st response you said that "You have to expand the column for this factor in the design matrix to the columns in the model matrix". Can you elaborate on this point, please.
Thank you