Before-Analysis: Check for any anomaly data
To be added into the survey:
Check whether 5 questions in total for each person (Should be 0 if no error)
## [1] 0
Following should be 100 if no error
## [1] 1000
Following should be 100 of 5s if no error
## .
## 5
## 1000
Following should be 100 of 0s if no error
## .
## 0
## 1000
## .
## 0
## 1000
Check whether each video gets roughly balanced assignment
## chipT1 chipI1 dT1 dI1 hfT1 hfI1 nT1 nI1 aT1 aI1 sgT1
## 254 253 259 249 252 247 253 258 253 248 247
## sgI1 rmT1 rmI1 crT1 crI1 lT1 lI1 hT1 hI1
## 232 259 252 252 257 254 245 240 236
Check whether all participants have 35 responses
## .
## 35
## 1000
Check how many influencer ads (out of 5) each person sees
## # A tibble: 6 × 2
## n count
## <dbl> <int>
## 1 0 37
## 2 1 162
## 3 2 305
## 4 3 313
## 5 4 149
## 6 5 34
Questions below were asked in this specific order even though the coding of the question has different order.
1. After watching this video, how positive or negative are your feelings towards Domino's?
5. How would you rate the production quality of video?
4. How much information did the video give you about the product/service featured in the video?
2. How trustworthy do you consider the message in this video?
6. To what extent do you think __[product]__ actually has the properties the video claims?
8. How connected do you feel towards the person speaking in the video?
7. To what extent do you think the person speaking in the video actually believed what they were saying?
3. Based on what you saw, how likely are you to purchase the product/service featured in the video?
Outcomes here are purchase intention, trustworthiness, and feelings towards the brand.
Data is in panel format, thus each person has multiple records. Filtered the panel data to only the three outcomes as shown above (should have equal weights under this setting).
Overall Treatment Effect
By Brand
By Number of Influencer Ads Seen
Note that there are 2 participants who saw all influencer ads and 0 participant who saw all traditional ads. For the 2 who saw all influencer ads, we cannot compare the treatment effect given that there is no traditional ads assignment.
Implication and Warning: This analysis might be ill-informed given the imbalance in the number of ads seen. Maybe should consider showing equal number of influencer and traditional ads to each participant.
Outcome here is the average of purchase intention, trustworthiness, and feelings towards the brand.