I have trouble interpreting the data from my survey project. Help!!!

zarunamuhaha

New member
Joined
Feb 25, 2017
Messages
1
I am trying to see how the appearance of the interviewer create response bias. My survey question is: "do you exercise on a weekly basis?"

I conducted a survey of 80 students using cluster sampling by choosing one room in each hallway, then four students in random from each. My control group of 40 students were asked by the interviewer in plain clothing, while the 40 left were asked while wearing sports attire.

The results of the survey are as follows:

a. Control Group: 65% said Yes (14 males and 12 females), 35% said No (3 males and 11 females)
b. Experimental Group 60% said Yes (18 males and 6 females), 40% said No (3 males and 13 females)


Since the interviewer in the experimental group was dressed in sports attire, we expected the students to be more compelled to say that they do exercise--but that is not that case. Am I missing something here? Or should I just conclude that wearing sports attire did not adversely affect student responses? In fact, there were 5% less people who said yes in the experimental group--would that small a number be just attributed to chance variation?
 
You are missing a significant difference.

You interviewed 17 males in the control group and 21 in the experimental group. Those who said "no" we're 3 in each group, but the percentage of nay-sayers was indeed lower in the experimental group as you expected. Whether that difference is statistically significant must be tested, but the evidence does not contradict your theory with respect to males.

You tested 23 females in the control group and 19 in the experimental group. 6 females in the control group said "no" (just over 25%), but 13 (around 67%) said "no" in the experimental group. That is evidence that does contradict your theory with respect to females though whether it confirms the contrary theory will depend on tests of statistical significance.

There may be a difference in reactions of males and females, but you might want to look at differences in the sex of interviewers as well. Although your samples will probably be too small to tell, you could also look at a group defined by interviewers of the same sex as the interviewed versus a group defined by interviewers of the opposite sex as the interviewed. A person might have different sub-conscious reactions if interviewed by a fit young person of the same sex dressed in gym clothes than if interviewed by a fit young person of the opposite sex in gym clothes. In other words, it is not clear to me that, in the experimental design, consideration was given to variables with obviously potential relevance.
 
Last edited:
Top