We select 100 people from the same representative population and ask who they will vote for. After conducting the survey, we collate their results and calculate the mean and standard deviation of voters over the five surveys.

After comparing our surveys, we realize that the population sampled is not similar in the five surveys. One selected the elderly, while the other one selected primarily the young people. To account for these potential biases, we decide to pool all our data and regress the potential factors using a stepwise regression model. We identified the following factors: gender, age, social networking activity (hours per day) social networking activity and in which state people reside. Which model should we use linear, logistic or Poisson and why?