# P for trend

#### Furby2020

##### New Member
Hi, how do I calculate the P for trend of crude incidence rates? Below is an example from a report. CIR=crude incidence rate in per million population. Thanks.

#### hlsmith

##### Less is more. Stay pure. Stay poor.
A pvalue for trend is kind of vague and doesn't tell you the magnitude or direction of the trend. Why not fit a Poisson model and include year as a predictor, so you can actually get an estimate with a precision value?

#### Furby2020

##### New Member
A pvalue for trend is kind of vague and doesn't tell you the magnitude or direction of the trend. Why not fit a Poisson model and include year as a predictor, so you can actually get an estimate with a precision value?
Dear hlsmith,
Thanks for your suggestion. Had tried log-linear regression using the ln(rate) regress on year and got the same results as the report.
Source | SS df MS Number of obs = 9
-------------+---------------------------------- F(1, 7) = 0.17
Model | .022463524 1 .022463524 Prob > F = 0.6922
Residual | .923709053 7 .131958436 R-squared = 0.0237
Total | .946172577 8 .118271572 Root MSE = .36326

------------------------------------------------------------------------------
ln | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
Year | -.0193492 .0468968 -0.41 0.692 -.1302425 .091544
_cons | 40.97742 94.40328 0.43 0.677 -182.2509 264.2057

------------------------------------------------------------------------------

For your suggestion, do i put the dependent variable as incidents (no. of cases) and include an offset in the model using the size of the population?

There are many methods to P for Trend. This is adapted from a SAS article. I was thinking cochran-armitage trend test is also applicable?