How do I get a p-value when dong proc means?

[gtucke1]

Hi,

My IV's are:

Spousal Status (0=spousal, 1=non-spousal)

Race (0=African American/Black, 1=White)

Gender (1=female, 0=male)

The DV is total % unmet needs

Conducted proc means to get the mean and SD of the total % unmet needs for the following: 

Female

Male

African American/Black

White

Spousal

Non-Spousal

Spousal Male

Non-Spousal Male

Spousal Female

Non-Spousal Female

Spousal African American/Black

Non-Spousal African American/Black

Spousal White

Non-Spousal White

Spousal male White

Non-spousal male White

Spousal male AA/B

Non-Spousal male AA/B

Spousal female White

Non-spousal female White

Spousal female AA/B

Non-spousal female AA/B

This is an example of one of the proc means:

Title "Non-Spousal Female Black";
proc means data = final.data N MEAN STDDEV;
var per_jhdcna_unmet_cg;
where spousalstatus = 1 and cg_sex = 1 and race_dich_cp = 0;
Run;

Now I want to find the p-value for the following groupings:

Gender (male/female)

Race (AA/B, W)

Spousal Status (spousal/non-spousal)

Spousal Status by Gender

Spousal Status by Race

Spousal Status by Gender by Race

Is there a test I can do to capture the M, SD, and a p-value? 

I'm a beginner, so I don't mind doing more steps if it is less complicated. 

[PaigeMiller]

p-value for test of what hypothesis? What are you comparing ... means or standard deviations or something else?

[gtucke1]

I'm comparing the means to determine whether male/female, non-spousal/spousal, etc. has the higher mean % of unmet needs. 

[PaigeMiller]

---

@gtucke1 wrote:

I'm comparing the means to determine whether male/female, non-spousal/spousal, etc. has the higher mean % of unmet needs. 

---

Normally, hypothesis tests compare if the mean of male and the mean of the female are equal, and standard hypothesis tests do not test which has the higher mean. A minor detail perhaps, but on the other hand maybe the standard hypothesis tests are not what you want.

In addition, you don't say what your sample size is, how many males and how many females. If the number is small (say <15 each), you might want to use PROC NPAR1WAY instead of PROC TTEST. If the distribution is terribly non-normal (percents being constrained between 0 and 100 inclusive), you might need even more samples in each category.

[Tom]

First step is to give your variables actual variable names.  So "total % unmet needs" might be UNMET (or whatever name is clear enough to understand and short enough to actually type into code).  "Spousal Status" could perhaps just be Spousal_Status.

To compare the means of two groups use PROC TTEST.

proc ttest data=have;
  class Spousal_Status;
  var UNMET ;
run;