Smoking cigarettes is noticeably prevalent on the Skidmore campus, despite the recent change in the smoking policy.. As non-smokers, we were curious about the motive for smoking, whether it be associated with stress or socialization, for example. Thus, in our Intermediate Data Analysis class, MA276, we were interested in finding out the best predictor of the probability that a Skidmore student would smoke cigarettes.
We created a survey on Survey Monkey and shared it on Facebook. We hypothesized that a student’s family history of smoking, along with GPA and gender might be used to predict whether a student smokes more than two cigarettes a week. Eighty-six students, mostly from the sophomore class, responded to the survey.
Prior to using statistical methods to determine the best predictor of our response variable, we assumed that male students smoke more than female students. This was completely drawn from our experience on campus. We also thought that family history of smoking would increase the probability of smoking. Finally, we assumed that lower GPAs would be associated with higher probability of smoking.
Drawing on Rstudio, a statistics program, we found that family history significantly correlates to whether a student smokes, while gender does not. The probability that a student with a family history of smoking smokes is 40 percent while a student having a GPA of 3.5 correlates with a 4 percent probability that he or she smokes. A student with a 4.0 GPA is predicted to have close to 1 percent probability of smoking.
Because the survey was open to Skidmore students, with the exception of the Class of 2018 as first-year students do not yet have a GPA on the files, we believe that our model avoided sample bias. However, one may assume that randomness does not hold because we shared it on our Facebook walls, which would limit the respondents to only our friends.
We conclude that the best predictor for the probability that a Skidmore student smokes is family history. GPA is also a good predictor, while gender is not. Despite these results, a population of 86 is not entirely representative of the Skidmore student body, so these conclusions should be taken with a grain of salt.