Suppose a researcher wants to determine if talking on the phone causes more pedestrian accidents than listening to music. The researcher recruits a group of university students and first has them fill out questionnaires about how often they use mobile devices while walking and how exactly they use them. He then divides the groups based on their answers. He asks those who reported that they almost always walk down the street listening to music in one group, and those who almost always are talking or texting on their phone in another group. He omits all the others with more moderate behaviors from further study. He asks all the participants to continue their behavior (walking while listening to music or talking/texting) over the course of the week and record any difficulties they have, including being honked at by cars, bumping into objects or people, near misses with cars or bikes, and any accidents. In one week, 80% of those in the music group returned and were asked to fill out a form that estimated the number and type of traffic incidents that participants experienced occurred over the previous week. Only 45% of the participants in the cell phone group returned for follow-up and the researcher conducted a face-to-face interview to assess how their behavior impacted their safety.
Identify and explain the threats to internal validity in this study.