i'll be blunt. this is not for amateurs. this is difficult work even for professional, degreed statisticians since you are looking for very subtle affects.

secondly, garbage in, garbage out. if a database was full of useful info that might lend itself to statistical analysis, it's been examined. if a database hasn't been examined, it's very likely it isn't useful.

thirdly, you are playing with fire. if you report a correlation and it proves to be false, think of the panic you would create.

that's why highly trained professionals do this work because if done wrong it can have very bad consequences.

with new biochemical biomarkers being identified and sites outside the brain such as the gut, olafactory cells, saliva glands, etc. having lewy bodies, with changes in sleep patterns, constipation, smell possibly being pre-motor indicators, with being able to create neurons from skin cells from pd'ers upon which you can cheaply and quickly test chemicals and other stresses, with ongoing studies which are following thousands of patients around the U.S. at risk / not at risk for PD for the first signs of pd, i think population studies aren't going to tell us much more. researchers are thinking some cases of pd are started in the womb by predisposing one to environmental stresses. if that's the case, then what causes pd in one patient might be totally different than in another if dozens of genes are involved and are affected by the environment in the womb or early childhood.

then with DNA analysis becoming very cheap, it will be possible to go beyond just identifying a few major and rare genotypes that give you a significant probability of getting pd to identifying more subtle genotypes so you can find out in your 20's what your chances of getting pd are assuming the causes are so widespread in the environment we can't easily avoid them. that is the goal of 23andme and other companies and to link their research in identifying relevant genotypes to drug development. and as far as reporting your results, consider the implications of genetic testing for just one person. 23andme reported to me that i had an allele for the BRIC1 gene which slightly increased my chances of getting breast and prostrate cancer since i'm a male. i told my sisters about this. one sister new a lot about this gene and was carefully getting exams. the other sister knew very little and totally freaked out. your're playing with fire if your're recruiting amateurs to examine databases. if you are asking people to report small clusters which might not be picked up in a big database, then go for it. the irony is i would think asking members of this board to participate in a clinical trial like the following would be far more useful than to as them to be amateur detectives.
https://foxtrialfinder.michaeljfox.org/trial/2708/

not only do you have to be a statistician to know if a database is normally sampled, i.e., does it represent a normal distribution, but is the data collected in the time/geographic window when the "affect" occurred? pd likely started in our 20's, yet most data you are examining is from data collected by people over 50. how do you "lock in" what event might have caused pd if it is incredibly subtle?
as far as smoking, tobacco has been shown to be neuroprotective as has tea and coffee.

i suggest you watch this webcast which does a good job at explaining pd starts way before motor systems occur and what is being done to find causes.
http://www.theparkinsonsgroup.com/webcasts.asp
the last 2 are the best but they are all interesting.

imho, bottom line, if you don't have experience in doing these types of analyses AND DEFENDING the results as to the epidemiology , it's a waste of time. if you just want to report interesting correlations/statistics as to drugs, costs, trends - is pd increasing/decreasing somewhere, etc. that would be very useful.