I am doing a research project for school where my dependent variable is "a response to a tweet" sentiment score (i.e. any value from -1 to 1-where 1 is completely positive, -1 completely negative, etc.). My predictor variables are are whether or not the "original tweet" mentioned a specific term (so 0=did not mention, 1=did mention). There is also another variable "the original tweet sentiment score" so again (any value between -1 to 1).
In addition to this, I am trying to figure out when the "original tweet" mentions a specific term (i.e. those with a 1 (did mention), whether that term is in reference to dark days (i.e. 1=yes, 0=no) or light days (1=yes, 0=no) and how this relates to the sentiment of the response to a tweet.
My question is does anyone know what test I would use for this? The only thing I can think of is the bayesian hierarchical regression model with proc genmod, they give several distributions but none seem to match (binomial, gamma, geometric, igaussian, multinomial, negbin, nromal, poisson, zip). Further I'm not sure how to identify a link function for a -1 to 1 dependent variable.
Here is my sas code for more information
proc GENMOD data=work.tweets; model RESPONSE_TWEET_SENTIMENT= WORD ORIGINAL_TWEET_SENTIMENT / DIST=POISSON LINK=PROBIT; BAYES; RUN;
Any help you can provide would be great in terms of should I use a Bayes Hierarchical regression? What might be the associated link and dist? and also is Gibbs sampling appropriate for this?