topic Re: Trends over time in Statistical Procedures

Trends over time

EMB — Sat, 24 Jun 2017 16:59:56 GMT

Hello,

I wish to study trends in age over time in 3 large samples (>1000 individuals each) from the same population and get p for linear trend.

Samples are independent and data are available for 4 time periods separated by equal time-intervals.

Would the following be correct? Any other suggestion?

proc glm;
class (time period);
model AGE=(time period);
contrast 'linear' (time period) -3 -1 1 3;
run;

Thanks!

Re: Trends over time

PaigeMiller — Sat, 24 Jun 2017 17:48:40 GMT

@EMB wrote:

Hello,

I wish to study trends in age over time in 3 large samples (>1000 individuals each) from the same population and get p for linear trend.

Samples are independent and data are available for 4 time periods separated by equal time-intervals.

Would the following be correct? Any other suggestion?

proc glm;
class (time period);
model AGE=(time period);
contrast 'linear' (time period) -3 -1 1 3;
run;

There are a few things you haven't told us, such as:

For each individual, are the time intervals exactly the same?
How can age be a response variable? If you measure the idividual's age at four different time periods ... isn't the individual's age already a known function of the age at the first time interval, plus the delta between the other intervals? What is the source of randomness in this data?

Re: Trends over time

EMB — Sat, 24 Jun 2017 19:10:58 GMT

The study includes incident cases of a specific disease, and data are collected at defined periods, each period includes different participants. As part of the descriptive analysis I would like to look at trends over time in age of the diagnosed individuals.

Re: Trends over time

PaigeMiller — Tue, 27 Jun 2017 17:39:33 GMT

@EMB wrote:

As part of the descriptive analysis I would like to look at trends over time in age of the diagnosed individuals.

I am totally not understanding this part.

Trends of what? Your sentence (and code) implies you want to find the trend over time of AGE. That cannot be right. Your age increases 1 year for every 1 year of change in time. No statistics or regression or p-values needed.