Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: zz-x2580
MSBA 7025 Group Project
You are asked to analyze the simulated data (project_data.csv) for an experiment.
Except for the following two differences, the experimental design is very much the
same as the one described in the paper (Huang, Shan, et al. "Social advertising
effectiveness across products: A large-scale field experiment." Marketing Science
39.6 (2020): 1142-1165.)
1. The randomization was triggered by the condition: there were at least one
organic likes from the friends of the ad viewers at the first ad impression.
2. There are only two conditions: Control - Show 0 like, Treatment - Show 1 like.
3. The sample size is smaller.
Please apply as many methods that you learned from the course as you can to
analyze the data and obtain insights for decision-making. Make sure to satisfy the
assumptions for all the methods that you apply. You may conduct the following but
not limited to the subsequent analyses.
1. Form hypotheses
2. Sanity checks
3. Power analysis
4. Compare means (or other summary statistics) across different variants.
• Statistical tests (t-test, z-test, bootstrap, etc)
• Correctly estimate the variance
• Try to improve sensitivity
5. Analyze the heterogeneous treatment effects if necessary.
6. Summarize your results
7. Inform product strategies for WeChat Moments Ads based on your results.
Requirements:
a. Please clearly describe your analyses and demonstrate the results in tables
and figures.
b. You can refer to the paper, Huang et al. 2021, but you are not allowed to copy
from it directly.
c. Your grades will largely depend on how well you understand the concepts,
theories, and applications of the methods taught in the course.
Data Schema:
User: User ID
Adid: Ads ID
Week: the first, second, or third week
Expid: Variants ID, 0: Control Group, 1: Treatment Group 1
If_click: an indicator for clicking ads at first ad impression of users (ad viewers)
Real_like_cnt: # of organic likes created by friends at users’ first ad impression
Category: Product category of ads
Brand_effect: big brands or not of ads
Experience: an indicator for experience goods of the products advertised in the ads
Status: an indicator for status goods of the products advertised in the ads
User_age: age of users (ad viewers)
User_gender: gender of users (ad viewers)
User_city: city class of users (ad viewers)
User_degree: the number of WeChat friends of users (ad viewers)
Friend_age*: age of the friend who created the first organic like for a (user, ad) pair
Friend_gender*: gender of the friend who created the first organic like for a (user, ad)
pair
Friend_city: city* class of the friend who created the first organic like for a (user, ad)
pair
Friend_degree*: the number of WeChat friends of the friend who created the first
organic like for a (user, ad) pair
user_sns_like_cnt - #likes that a user created on moments last month
user_sns_comment_cnt - #comments that a user created on moments last month
*This friend’s like was shown in the ads for the (user, ad) pairs in the Treatment
Group but was hidden from the ads for the (user, ad) pairs in the Control Group.
Note:
The objective of the group project is to let you practice the methods learned from the
course systematically and comprehensively. The research paper is only a reference.
You don't need to replicate all its analyses. Some of the analyses in the paper are
beyond the scope of our course.
You shouldn't expect the experiment to be perfect. If you find some problems, you
can specify your reasoning. For example, if you think there is a problem with
randomization checks, please describe it based on your analysis. Also, please finish
all the other analyses assuming the randomization is valid.
