about the cumulative risk graph

[twotwotwo]

Hi,

I have the data like the following. disease=1 means having the disease; disease=0 means no disease.  Can I draw the cumulative risk graph?  The x-axis is the response (continuous variable), y-axis should be the cumulative risk (probability??) of having the disease. How can I do it?  Thanks!

data test;

input response disease;

cards;

45.6   1

190     0

34.5    1

89       0

56       1

45       1

58       0

.......

........

;

[Reeza]

Possibly dumb question, but if you have the disease isn't your probability of cumulative risk of having the disease 1?

If you're getting the cumulative risk data from a proc some have the ability to plot them from within the procedures.

If not you could use Proc SGPLOT.

[twotwotwo]

Thanks for your reply.  Probably I should ask how to calculate the cumulative risk?  Do you know any proc can do it?  Thanks!

[Reeza]

I'm not sure what your data represents.

My guess is it looks like survival data based on your question and then you'd use proc lifetest or proc phreg.

0/1 is your censor variable and response is the time to either having the disease or not, if its not a time measurement then reply back.

Try proc lifetest and check the options.

[twotwotwo]

Thanks Reeza!  It is not a survival data.  The response is a measurement for a patient.  For this kind of data, how can I do it? Thanks!

[Reeza]

I think you've posted before about this and again the question is still too general to give any comments without more specifics.

If you can't post more details then I recommend finding some similar published articles and seeing what method they used, ie logistic, survival, and then more help could be provided around the specific procedure rather than a general question.. 

[twotwotwo]

Hi, Reeza,

I did not post this question before.  What other information do you want for this question? 

[Reeza]

What's your response variable?

What's your research question that you're trying to answer?

Is this hw?

[twotwotwo]

My response variable is one measurement for the patient, like the blood pressure, height, BMI....

My research question is to see: with the measurement differing, what is the risk for a person to get the disease (disease=1). 

Please let me know if you have other questions.  Thanks!

[Reeza]

Ok, I guess with that type of data, I'd be looking at a logistic regression and odds of getting the disease.

See  Example 51.13 Complementary Log-Log Model for Infection Rates

In the SAS 9.2 Docs.

The last table plotted should help give you what you're looking for.

[twotwotwo]

Thanks! I then found that the effectplot is what I want....