Variable Selection using Stepwise Regression with Application to MyTherapistMatch.com
MyTherapistMatch.com is a website that connects patients with therapists who are deemed to be well suited to patients' personality and mental health needs. The primary tool for this patient-therapist matching is an online questionnaire that is completed by users visiting the site. Only approximately ten percent of visitors to the site complete the questionnaire, ostensibly partly due to the excessive length of the survey. The purpose of this thesis is to identify the items on the questionnaire that have the greatest impact on how much users interact with the website and reach out to their therapist matches. This is done with linear regression techniques, including ordinary least squares regression and stepwise regression. According to the regression analyses, the single greatest determinant of user activity is whether the respondent lives in or outside of California. The proprietors of the website may use the results of this analysis in choosing a meaningful subset of items that forms a shorter questionnaire.