which SAS test should I use to show change over time?

[Angmar]

Hi,

Thank you for your help in advance. 

I'm unsure which test I can use in SAS for the following:

I need to demonstrate that increasing temperature over time leads to an increase in algal blooms over the same time period (the variables would be temperature (interval), and year and month, and outcome is algal bloom. On top of this, I need to determine whether cases of human illness have increased due to the increase in algal blooms. A leads to B which leads to C, etc. I won't have more information than this, so I'm hesitant to move into real tests of association - I think descriptive tests are all I can run.

Any advice would be greatly appreciated.

[PaigeMiller]

A better and more precise problem statement would help.

increasing temperature over time leads to an increase in algal blooms ...

So which is it that is increasing that you want to work with? It would seem to me that temperature is the variable related to increasing algal blooms, but you specifically say time in the title.

... over the same time period

Not sure what this means at all, or how it relates to the previous snippet of your problem statement.

I won't have more information than this, so I'm hesitant to move into real tests of association - I think descriptive tests are all I can run.

Oy vey. How are we supposed to help, if you don't even have more information about what the problem really is?

So I doubt we can state a test to do this with this little information ... but let me ask you this ... have you even plotted the data? Sometimes, that's enough to get you started, and sometimes that enough to demonstrate that there's nothing there — or to demonstrate a really strong effect.

[Epidemiologist1 · Accepted solution]

If I understand what you are asking, you want to test for trend over time. You can do this several ways. If you have categorical data, use proc freq with a trend option. If you are testing a continuous variable, you can do a regression analysis and look at the stat. sig. of the slope variable. Same with human illness.

When comparing illness to algal bloom, again, you can do a t-test if you have continuous data for each. Similarly, you can do a 2 way table for categorical variables (proc freq). You can never "prove" that the 2 are causally related-only show association, but it will give you some idea if there is a similar pattern. 

[PGStats]

"demonstrate that increasing temperature over time leads to an increase in algal blooms"

hypothetically

you can show that there is an increase in temperature (air, water?)

you can show that there is an increase in algal blooms

you can show that there is an increase in fertilizer usage

etc.

... But to show that two monotonic trends are related you need strong evidence. Look at the Granger test of causality in PROC VARMAX (SAS/ETS) and at the new CAUSALMED procedure.