topic Re: simple linear regression for yearly trend with sample weights in SAS Programming

simple linear regression for yearly trend with sample weights

Claw — Tue, 08 Jul 2014 18:28:49 GMT

Hi All,

I am beginner user for SAS. I have large pubic data with multiple variables including sample weights. I am trying to do simple linear regression to check for yearly trends in mortality.

my question will be "trends in mortality over time".

my extracted data has variables 'year', 'died' (0=alive, 1=died) and 'weight'(sample weight). I am planning to use proc surveyreg but can't seem to get SAS code right. Please help.

my data looks like as follow (around 5,00,000 observations in dataset)

year died weight

2002 0 4.3383

2002 1 5.1405

2003 1 5.2034

2011 0 5.9179

2011 1 4.9550

I am looking for something as follows (below is the example from the different article and I am looking for similar analysis)

Thanks in advance

Claw

Re: simple linear regression for yearly trend with sample weights

ballardw — Tue, 08 Jul 2014 21:18:43 GMT

post some examples of the code you tried and any associated error messages.

Re: simple linear regression for yearly trend with sample weights

Claw — Wed, 09 Jul 2014 02:23:22 GMT

i tried the following code. It seems completely wrong because 'died' is categorical variable.

proc surveyreg data=allibd;

model died=year / solution;

weight discwt;

run;

I need to know if there is upward / downward/ neutral trend in percentage of individuals from 2002 to 2011.

Re: simple linear regression for yearly trend with sample weights

ballardw — Wed, 09 Jul 2014 14:32:15 GMT

It looks like you may need to summarize the data to get PERCENT of the category of interes to model. Don't forget to accumulate the weights if so.

Re: simple linear regression for yearly trend with sample weights

SteveDenham — Wed, 09 Jul 2014 15:07:25 GMT

Rather than SURVEYREG, you should look at SURVEYLOGISTIC as a method. The examples in the documentation reference more than two categorical levels, but it should not be a problem to reduce the complexity in those examples to what you are doing.

Steve Denham