[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
st: RE: Secular trends
This sounds, deep down, to be a logit model problem
with -cc- as response. Or is it a covariate?
The first question is what the covariates should be.
You also need to consider quite what independence
or dependence assumptions you are entertaining. It
seems sensible to use a panel structure here for wards.
As a point of terminology, your tribe may differ, but tribes
whose literatures I read regard (secular) "trend" as long-term and
as distinct from "seasonality". That doesn't rule out the
seasonal effect varying secularly.
A naive model would have (a) month of year and
(b) year as covariates. If you find effects under
(a) or (b) you need to consider whether the model
deserves a time series flavour.
McFarland, Lynne V
> I have a data set with variables below:
> cc (binary 0, 1 for infected or control)
> beddays (number of occupied beds per ward per month)
> ward (ward patient was admitted to)
> onset (month of infection onset)
> My two questions are:
> 1. Does the incidence density of this disease vary significantly by
> month? (is there a seasonal trend?)
> 2. Does the seasonal incidence density vary significantly BY
> WARD ? (ie,
> are there "hot wards").
> Can someone help me writing the commands for these two
> questions? I got
> bogged down in the
> manuals and am brain fried!!!!
* For searches and help try: