topic Question about PROC CORR in Statistical Procedures

Question about PROC CORR

Relax — Thu, 12 Mar 2020 13:45:25 GMT

Hello All,

I have some JMP codes like below:

dt_a = dt_b << Multivariate(
Y( :Name( "A" ), :name( "B" ), :name( "C" ) ),
estimation method( "Row-wise" ),
invisible
);

I would like to replicate the results of these JMP codes in SAS. I know PROC CORR may give the correlation, but I don't know how to use "Row-wise" method in this procedure. Can someone help me?

PS: I am sorry if I posted this question in the wrong location.

Re: Question about PROC CORR

PaigeMiller — Thu, 12 Mar 2020 14:20:22 GMT

Please explain what JMP is doing when you say "row-wise"

Re: Question about PROC CORR

mkeintz — Thu, 12 Mar 2020 14:23:41 GMT

By "rowwise" do you mean your want correlations using values in the same row?

Second question: what do you want to do about missing values? If you want correlations of A, B, and C, when a row occurs with missing B, but valid A and C, do you want that row to still contribute to generating the correlation of A and C?

Re: Question about PROC CORR

Relax — Thu, 12 Mar 2020 14:32:25 GMT

Thanks for replying. Row-wise estimation is used for data tables with no missing values and calculates the correlation of the variables in the same row.

Re: Question about PROC CORR

PaigeMiller — Thu, 12 Mar 2020 14:34:18 GMT

@Relax wrote:

Thanks for replying. Row-wise estimation is used for data tables with no missing values and calculates the correlation of the variables in the same row.

No missing values ...

proc corr data=mydatasetname;
    var a b c;
run;

Re: Question about PROC CORR

Relax — Thu, 12 Mar 2020 14:35:57 GMT

Thanks for replying.

Yes, I want correlations using values of A, B and C in the same row.

Row-wise estimation is used for data tables with no missing values, so if a row occurs with missing B, but valid A and C, I will delete this row.

Re: Question about PROC CORR

PaigeMiller — Thu, 12 Mar 2020 14:39:26 GMT

@Relax wrote:

Thanks for replying.

Yes, I want correlations using values of A, B and C in the same row.

Row-wise estimation is used for data tables with no missing values, so if a row occurs with missing B, but valid A and C, I will delete this row.

This is different than your previous explanation.

Try this

proc corr data=mydatasetname nomiss;
    var a b c;
run;

Re: Question about PROC CORR

Relax — Thu, 12 Mar 2020 14:44:02 GMT

Thank you, Paige. I appreciate your help.