Regional analysis of disease risk factors

#1
I'm working with a large data set looking at climate variables e.g maximum, minimum temperature etc to see whether these could be risk factors for the regional variation for a sheep disease in 5 different regions within a country. The dependent variable is disease count (number of cases).

Climate variables values are given per region rather than per post code e.g the whole of region 1 sheep farms (North) with 200 farmers responding to questionnaire the temperature value = 15, region 2(south) temperature=16.

The values for the outcome (number of cases/sick sheep) has been given per farm/post code.

My main problem is the small margin of variation between independent variables seem too be too small making it harder to model them with the outcome or come up with best method of analysis.

Is the statistical method I could employ to help work out the association between climate variables and number of disease cases given such small margins when it comes to independent variables.

Any advise will be much appreciated and thanks in advance
 

rogojel

TS Contributor
#2
hi,
is there a regional variation that needs to be explained? And if there is, could it be attributed to other factors - such as different races etc.? On a regional scale, as you observed,mthe climatic differences might just be too small to have any impact.

regards
 
#3
Yes, regional difference needs to be explained too. the data has around 15 climate variables which all vary between regions some with very small variations. What would be the best statistical approach when it comes to looking into these factors as predictors of the varied disease counts between regions.

Thanks
 

rogojel

TS Contributor
#4
hi,
I would simply build a model that explains the regional differences without singling out the climatic variables. If in the best model to explain the regional differences you have climatic variables, fine - if not, that is fine as well. It seems to me, that you want to find a model to fit the preconception that climatic variables have an effect - I guess that would be the wrong way to use statistics.

regards
 
#5
Thanks for the prompt response Rogojel.

I have to include the climate variables as these are the only predictor variables - the main aim is to see if these variables predict or associated with the number of disease cases (counts) reported in the 5 regions, thus a model has to include the climate variables. Tried different multivariate models e.g poisson, quasi-poisson without much success possibly due to the small variations between variables. Any specific statistical approach which you would suggest?

very many thanks