Proc tabulate with several vars and several formats for descriptive statistics

[dariatri]

Hi,

I'd be happy to receive help 🙂

That's my code: 

proc tabulate data=people (where=(SAFFL='Y'));
var AGE weight height BMI;
table AGE='Age [years]' weight='Weight [kg]' height='Height [cm]' BMI='BMI [kg/m2]', (mean*f=8.1 std*f=8.1 (min median max n)*f=8.)/box='Parameters';
run;

I only need the BMI var statistics to be in 8.1 format and the rest in 8.0. I'm not able to do that. The only way I can think of is to create a separate table for BMI and somehow combine them.  Is it possible to do it in the above code?

Thank You!!

[pink_poodle]

You might need to associate the 8.1 format with the variable name (BMI) in addition to the statistic keyword.

Use the asterisk (*) operator to associate a statistic keyword with a variable. The N statistic (number of nonmissing values) can be specified in a dimension expression without associating it with a variable. [1]

This might work. 

table 
AGE='Age [years]' 
weight='Weight [kg]' 
height='Height [cm]' 
BMI='BMI [kg/m2]', 
(BMI*mean*f=8.1 BMI*std*f=8.1 (min median max n)*f=8.)
/box='Parameters';
run;

Also, try enclosing the stat names in quotation marks:

 Statistical keywords should be enclosed by single or double quotation marks to ensure that the keyword element is treated as a statistical keyword and not treated as a variable [1].

Reference

[1] TABULATE Procedure: keywords for statistics in dimension expressions

[ballardw]

---

@dariatri wrote:

Hi,

I'd be happy to receive help 🙂

That's my code: 

proc tabulate data=people (where=(SAFFL='Y'));
var AGE weight height BMI;
table AGE='Age [years]' weight='Weight [kg]' height='Height [cm]' BMI='BMI [kg/m2]', (mean*f=8.1 std*f=8.1 (min median max n)*f=8.)/box='Parameters';
run;

I only need the BMI var statistics to be in 8.1 format and the rest in 8.0. I'm not able to do that. The only way I can think of is to create a separate table for BMI and somehow combine them.  Is it possible to do it in the above code?

Thank You!!

---

Easiest may be to use two table statements (you can have many table statements in proc tabulate).

proc tabulate data=people (where=(SAFFL='Y'));
var AGE weight height BMI;
table AGE='Age [years]' 
      weight='Weight [kg]' 
      height='Height [cm]' ,
     (mean*f=8.0 std*f=8.0 (min median max n)*f=8.0)
     /box='Parameters';
table BMI='BMI [kg/m2]', 
     (mean*f=8.1 std*f=8.1 (min median max n)*f=8.1)
     /box='Parameters';
run;

if I understand what you want.

The other approach would be to presummarize the data with a procedure such as Proc Means/Summary, possibly transpose and/or round the results as desiredthen use proc Print or Report to display the values with best. format