lag function to estimate risk differences

[Olivi]

Hello everyone,

I am calculating risk differences with the lag function, I have a variable exposure category (with 3 exposure levels), I want to calculate the difference based on the reference category, but I have not succeeded, any suggestions?
help please

[Olivi]

Hello everyone, I am calculating risk differences with the lag function, I have a variable exposure category (with 3 exposure levels), I want to calculate the difference based on the reference category, but I have not succeeded, any suggestions? help please.

I am using this code, there is no problem for the difference between category 2 with the reference, but the calculation of the third category is no longer done with the reference, rather with the previous category (which I understand is what happens with the lag function) is there any way to set a reference category?

data test;
set test;
by time exposure;
rd = risk - lag(risk);
rr = risk / lag(risk);
if first.time then rd=.;
if first.time then rr=.;
run;

[Patrick]

Please amend below HAVE sample data if it is not suitable for your problem.

Once you've created good enough sample data how should the desired result look like?

data have;
  infile datalines dsd truncover;
  input time exposure risk;
  datalines;
1,1,1
1,1,2
1,2,3
2,1,1
2,3,3
;

data want;
  set have;
  by time exposure;
  rd = risk - lag(risk);
  rr = risk / lag(risk);
  if first.time then call missing(rd,rr);
run;

proc print data=want;
run;

[Olivi]

I will try, thank you

[Kurt_Bremser]

What is the reference category?

Please also supply usable example data (in a working data step with datalines), and the expected result.

[Olivi]

The exposure category has 3 values (0, 1 and 2) and the reference category is 0.

Example data:

data test;

input exposure time risk:

datalines;

0   0   0.01413

1    0    0.0125

2   0    0.1077

0   1     0.0491

1    1     0.0838

2   1     0.0678

;

I want to estimate the risk difference based on the reference category using the lag function.

[Patrick]

Using your sample data how would the desired result look like? Providing such sample want data often helps a lot to clarify the requirements.

Not having such want data I'm not sure if I've got below formulas right - but the code should anyway give you the idea.

data test;
  input exposure time risk;
  datalines;
0 0 0.01413
1 0 0.0125
2 0 0.1077
0 1 0.0491
1 1 0.0838
2 1 0.0678
;

data want;
  set test;
  by time exposure;

  array risks {0:2} 8 _temporary_;

  risks[exposure]=risk;

  if last.time then
    do;
      rd = risks[0] - risks[1];
      rr = risks[0] / risks[2];
      output;
      call missing(of risks[*]);
    end;
  keep time rd rr;
run;

proc print data=want;
run;

Note: With a _temporary_ array the values get retained. 

[Olivi]

Thank you very much!