How I could get the correlation between the mortality and the clusters?

[Ana]

Dear All,

I'm new to this area and I need your help!

In order to study the relation between mortality [ 0 (survived) or 1 (died)] that 68 strains produced and the test (I to X), I have applied an univariable logistic regression analysis with proc genmod. So far so good.

proc genmod data=data descend;
    class Strain Test;
    model Mortality= Strain Test/ dist= binomial link= logit type3 noint aggregate;
    lsmeans Strain Test/diff;
run;

Now, I need help with the following. I want to assess the associations between the mortality with the ERIC-PCR results (clusters for each strain, from 1 to 7) and one sequence type (ST) of the strains (binomial variable; 0 = absence and 1=presence).

How I could get the correlation between the mortality and the clusters?

To the association between mortality and the sequence type (ST), I have attempted using the next procedures, and I have achieved the same result, but I am not sure if it is right:

proc genmod data=data descend;
    class ST;
    model Mortality= ST / dist= binomial link= logit type3 noint aggregate;
    lsmeans ST /diff;
run;

proc anova data=data plots(maxpoints=none);
    class ST82;
    model porcent*=St82;
    means ST82/ Bon;
run;
* porcent = number of subjects dead/number of subjects used per strain (in porcent)


proc ttest data=data;
    class ST;
    var porcent;
run;

I would appreciate some help with the correct syntax for this question...

Thanks in advance!

Ana

[pearsoninst]

Use

proc genmod data=data CORR descend;

[Ana]

Dear pearsoninst,

Thank you very much for your help.

Unfortunately, CORR is considered a syntax error and therefore it is ignored.

Do you have any other suggestions or idea?

[pearsoninst]

I am very sorry , i did not realize you are are using genmod . I i thought it is Proc REG . Corr works for that ..I have not idea on Genmod. Very sorry

[Ana]

Thank you all the same!
If you know, how I could get the correlation between the mortality (binary variable) and the clusters with the Proc REG is also well received.

[pearsoninst]

proc surveyreg data=newdata total=10;
   cluster Clustervariable;
   model yvariable=xvariable;
run;

and to find only the correlation use CORR Procedure.

[Ana]

Thank you for your suggestion dear pearsoninst!

I don't know the proc surveyreg but I will prove it!
About proc CORR, what would be the best correlation coefficient to analyze my data?

Thank you in advance!