Individual Assignment: Analyze an A/B Test
Introduction - For this assignment, you are to analyze data from an advertising image "pretest" conducted on Amazon Mechanical Turk (mturk.com).
Overview - Dulcet, a relatively new perfume brand wanted to know which image to use for a new campaign. The firm decided to test five different advertising images (listed below) to assess how consumer sentiment and get a sense of what might be more versus less effective. Many fragrances tend to promote feelings of confidence and sexiness, and Dulcet is no exception, however, the brand manager for Dulcet is unsure how far to push the envelope and is debating using more wholesome images versus more risqué images in the campaign. You have been tasked to help advise the brand manager with this decision.
Below are ad images from the pretest.Additionally, you will receive an Excel data file with responses from the "pretest" along with a copy of the questionnaire (Word doc). You will use these files to answer questions and develop insights. You will turn in responses to questions in this Word doc and the Excel file (showing the analyses and calculations reported in your responses in this document). You do not need to submit the questionnaire file, as this document will be unchanged.
Step 1: Research Design
Discuss whether a "pretest" was the best approach for the problem. Indicate whether you feel Dulcet was correct in selecting a "pretest" or if a "live" A/B test would have been more effective and provide support for this selection.
Step 2: Data Cleaning
Dulcet is interested in the opinions of women for this campaign. The A/B testing asked explicitly for women participants, however, some men completed the study. You need to eliminate the men from the data set prior to analysis.List the ResponseIDs below for the records you deleted from the data set prior to analysis.
You will also need to perform a calculation to transform height and weight into inches and then to calculate BMI. You should eliminate (replace with a ".") any height/weight responses that are outside of normal expected ranges (you must define this).
Write a short summary of what responses were affected and in what way (using ResponseID) to index any changes or adjustments you made.
Step 3: Descriptive Statistics
Using the cleaned data set, provide the mean value for each item in the tables below.
For BMI (table 2), you will need to calculate this separately. It is not provided for you in the data.
Table 1
|
AD/DV
|
ADRATE
|
ATTRCTV
|
SEXY
|
CLASSY
|
UNREAL
|
RECOG.
|
|
AD 1
|
|
|
|
|
|
|
|
AD 2
|
|
|
|
|
|
|
|
AD 3
|
|
|
|
|
|
|
|
AD 4
|
|
|
|
|
|
|
|
AD 5
|
|
|
|
|
|
|
Table 2
|
VARIABLE
|
MEAN
|
MEDIAN
|
MIN
|
MAX
|
|
AGE
|
|
|
|
|
|
GENDER
|
|
|
|
|
|
HEIGHT
|
|
|
|
|
|
WEIGHT
|
|
|
|
|
|
BMI
|
|
|
|
|
Step 4: Deeper Analysis
Dulcet is primarily targeted toward younger, single women. Define an age range you feel is appropriate, and then test each ad according to how single versus not single women respond. (Note: you will need to recode the data to perform this analysis). List the total number of observations (respondents) for each ad in the column "N" in the table.
Single
|
AD/DV
|
N
|
ADRATE
|
ATTRCTV
|
SEXY
|
CLASSY
|
UNREAL
|
RECOG.
|
|
AD 1
|
|
|
|
|
|
|
|
|
AD 2
|
|
|
|
|
|
|
|
|
AD 3
|
|
|
|
|
|
|
|
|
AD 4
|
|
|
|
|
|
|
|
|
AD 5
|
|
|
|
|
|
|
|
Not Single
|
|
N
|
ADRATE
|
ATTRCTV
|
SEXY
|
CLASSY
|
UNREAL
|
RECOG.
|
|
AD 1
|
|
|
|
|
|
|
|
|
AD 2
|
|
|
|
|
|
|
|
|
AD 3
|
|
|
|
|
|
|
|
|
AD 4
|
|
|
|
|
|
|
|
|
AD 5
|
|
|
|
|
|
|
|
Step 5: Recommendation
Make a recommendation on which ad Dulcet should use for it's target market and explain why. I would like for you to base your answer on the data (not your personal opinion). However, you do not need to conduct any statistical tests. It is fine to say Ad X was rated the highest in dimension A, or Ad Y was highly rated among dimensions A and B, or Ad Z had the best combined score along some metric you define. That said, I will award up to 5 bonus points for answers that appropriately uses statistical analyses as supporting evidence. Additionally, I will award up to 5 bonus points for additional analyses and investigations above and beyond what is asked specifically. That is, if you conduct additional analyses or deeper analyses and discuss those results in your recommendation, you can earn additional points.
Deliverables
You should turn in answers to the items outlined in Steps 1 - 4 above. You can answer within the text of this document. My text is blue. Your text should be in black to help distinguish responses from instructions.You will turn in both a copy of this document with your responses and the Excel data file showing the calculations supporting your answers above.
Attachment:- Assignment.rar