turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

Find a Community

- Home
- /
- Analytics
- /
- Stat Procs
- /
- Survival Analysis Question: How would I go about f...

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

11-06-2014 03:46 PM

Hello SAS community, I’m still stuck on this problem. I have looked further and have come to a standstill. Any help with be appreciated.

I have a homework problem where a sample of bone marrow transplants for Hodgekin's and NON-Hodgekin's Lymphoma. All patients had either Hodgekin's disease (HOD) or non-Hodgekin's lymphoma (NHL) and were given an allogeneic (allo) transplant or autogeneic (auto) transplant. This information is non-identifiable and taken from a graduate text book.

So, total there are four types of groups of patients. the first variable graft is used as a dummy variable . The variable, Score, is the pretransplant Karnofsky, Z1 and the variable, Wait, is the waiting time to transplant. This is unidentifiable data.

I have two methods of coding for type of disease (HOD or NHL) and type of treatment (allo or auto) below:

data problem8_5;

input trt disease time status z1 z2;

*first coding method: each 4 group as separate dummy variables;

if trt=2 and disease=1 then NHLauto=1;

else NHLauto=0;

if trt=1 and disease=2 then HODallo=1;

else HODallo=0;

if trt=2 and disease=2 then HODauto=1;

else HODauto=0;

*2nd coding method: type of treatment, type of disease, and interaction;

if trt=2 then auto=1;

else auto=0;

if disease=2 then HOD=1;

else HOD=0;

datalines;

1 1 28 1 90 24

1 1 32 1 30 7

1 1 49 1 40 8

with the following models, respectively:

proc phreg data=problem8_5;

class NHLauto(ref='0') HODauto(ref='0') HODallo(ref='0')/param=ref;

model time*status(0)= HODallo NHLauto HODauto/ties=breslow;

run;

*and;

proc phreg data=problem8_5;

class HOD auto;

model time*status(0)= auto HOD auto*HOD/ties=breslow type3;

contrast 'HOD auto vs NHL auto' HOD 1 -1 auto 0 1;

run;

I successfully found the p-value of the test that the survival rates of HOD auto vs. NHL auto are the same in the second PHREG procedure with the CONTRAST statement, which came back as the correct p-value=0.31, but just switching 'auto 0 1' to 'auto 1 0' does not give me the correct p-value for the two allo groups (which should be p-value=0.007)

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

11-06-2014 04:29 PM

I think my main question is how do you use the contrast statement exactly for this problem. I found the resource for the syntax, but I don't know how to interpret the data in order to code correctly.

I took a glance at the contrast statement from the SAS main website, but I still don't understand....:smileyconfused: