Dear allstaters
I have recently been invited to carry out statistical analysis for a project involving the assessment of the influence on meteorological conditions on onset of a particular disease over a period of about 7 years. I have 78 cases of onset together with their dates. However, on questioning my colleagues on how the data was collected, I discovered that there were several other cases (44) for the geographical region and time period of interest for which the data of onset was unknown.
Originally, I had planned to carry out a binary logistic regression on the data, where the meteorological conditions for days in which patients did not become ill or experience an incubation period were to be regarded as relevant to the outcome "no event". On learning about the 44 cases above, however, I felt that I was in no position to define a control, it being possible that at least one of my "no event" days could have coincided with a date of onset or an incubation day for at least one of these 44 cases. I felt therefore that I had hit a brick wall. Are there any more experienced medical statisticians out there who could kindly advise me as to what is the normal practise for the statistician in such circumstances. (I have been informed that by its very nature, epidemiological data is like that.)
I am eager to receive some ideas.
Many thanks in advance.
Margaret MacDougall
---------------------------------
ALL-NEW Yahoo! Messenger - all new features - even more fun!
|