Thread: Political Polls
View Single Post
Old March 7th, 2006, 12:09 AM  
Webbeardthepirate
Member
 
Join Date: February 14, 2006
Default

Your survey suffers from self-selection bias so its not valid satistically anyways. Pirate is perfectly good sexual orientations as far as sexual orientations go, which is to say not very far. Without a significant and randomized sample you can't really extrapolate to a larger population any ways. All you will really learn is about the individuals who choose to answer your survey. The evidence is anacdotal, and thus generally speaking irrlevent. If your goal is instead to get some idea bout some of the people whose posts you read oon an individual basis then you servey is fine for that. I wouldn't be compiling descriptive statistics or trying multivarient regressions based on answers to these questions.

RAND corporation, a non profit, is always doing work like this if you'd like to view their results of these sorts of survey's done on the grand scale. RAND research is generally reguarded as reliable among academics. But remember, descriptive statistics only show correlation, not causation. For causation you need to do a regression, and if your dependant variable is something like Liberal or Conservative then you have to do a binomial regresssion and the result is a number between zero and one and is much harder to show significant fit. I'd assume that you wouldn't be trying to show that liberal or conservative causes gender, or minority membership which is not supportable theoretically. Being liberal can't make you into a black woman or a white man, for example.
Webbeardthepirate is offline   Reply With Quote