Send the link below via email or IMCopy
Present to your audienceStart remote presentation
- Invited audience members will follow you as you navigate and present
- People invited to a presentation do not need a Prezi account
- This link expires 10 minutes after you close the presentation
- A maximum of 30 users can follow your presentation
- Learn more about this feature in our knowledge base article
stats project response bias
Transcript of stats project response bias
Given a recent study conducted at the University of London stating that women are more likely to be worse drivers than men, don’t you think that women’s car insurance rates should be higher than that of men?
c)No opinion Method Part 2 Survey #2:
Do you think that women’s car insurance rates should be higher than that of men?
c)No Opinion Design of Study 60 random subjects Description of Design Part 1 The random selection of 60 subjects, of which 30 are male and 30 are female, ensured an accurate sample that reflected our target population (adults and teenagers residing in Sugar Land). The blocking of the sample by their gender also reduced variation between the results (women might reply “No” because of their “loyalty” to their gender). The random assignment of surveys allowed us to see whether wording of the question can induce a response bias in those surveyed. Men Survey #1 (15 subjects) Men Survey #2 (15 subjects) Women Survey #1 (15 subjects) Women Survey #2 (15 subjects) Random Assignment Compare Responses Description of Design Part 2 The 60 random subjects residing in Sugar Land, were contacted through Facebook and email, so no personal contact was made in presenting the survey. The subjects were pulled randomly from a list of our contacts (ranging from teenagers to adults) compiled beforehand, in which every eligible person was classified by gender into two groups (male or female) and assigned a number from 001 to 200. We used a random number generator to generate 30 numbers from each group, to produce a total sample frame of 60 subjects Description of Design Part 3 This design represents the population of Sugar Land that can own and operate a vehicle (from teenagers to adults). This design accurately and most efficiently dealt with this survey, because very little room was left for unintentional biases through the complete randomization at every step of the design Question 1 (Biased) Results Yes No No Opinion Men Women 11 3 2 12 2 0 Question 2 Results Yes No No Opinion Men Women 8 1 5 13 2 Response Frequency Question 1 (Biased) Response Frequency Question 2 Conclusion Part 1 After conducting this experiment, we can conclude that response bias did truly occur, mainly due to the wording of the question that sharply contrasted the results from the biased Question 1 to the neutral Question 2. Conclusion Part 2 1 The main factor that can demonstrate this sudden change is the male response from the two questions, a decrease in nearly 27% acceptance of the question posed. The women response stayed nearly the same, hovering around 12/13 for both questions, demonstrating the response bias as well as each person's "loyalty" to their gender. Conclusion Part 3 We encountered very few problems, since our results were received in a timely manner from all the subjects and thus we were able to compile and analyze our data without any difficulties. One minor issue we faced was the difficulty in randomly selecting 60 acquaintances from an exhaustive list of our contacts online. Conclusion Part 4 We decided that this design is the most efficient way to conduct this study, and thus we aim to not change any aspect of it if we were to repeat it for future extrapolations. The one minor change that must be dealt with regardless of the experiment is the increase in the sampling frame (an increase in subjects) to accurately portray a generalized conclusion and allow the experimenter to make accurate future analysis from his data. Conclusion Part 5 We learned the importance of randomization in a survey/experiment, the need for a diverse and large sampling frame, and the significance of bias in the subconscious manipulation of survey results. The last point shows the uselessness of unverifiable polls and surveys, because at times, randomization and bias may not be accounted for and thus can eliminate the validity of any experiment.