'VPN mental models'

Download PDF

---

---

1) Have any data been collected for this study already?
No, no data have been collected for this study yet.

2) What's the main question being asked or hypothesis being tested in this study?
Our primary goal is to explore the relationship between exposure to influencer VPN ads and relevant (see 3) mental models.

We also will investigate the relationship between brand familiarity and exposure to influencer VPN ads.

3) Describe the key dependent variable(s) specifying how they will be measured.
We will measure mental models conveyed in influencer VPN ads through the following four scales:
- Threat mental models: combination of 6 likert items.
- Correct VPN mental models: combination of 6 likert items.
- Wrong VPN mental models: combination of 5 likert items.
- All VPN mental models: Combination of all VPN mental models questions (previous two VPN related scales plus two more items).

Brand familiarity is measured through Likerts for each of the top four most frequently advertised VPN brands.

4) How many and which conditions will participants be assigned to?
n/a

5) Specify exactly which analyses you will conduct to examine the main question/hypothesis.
The main analysis plan:
- We will run four two-lines analyses with the ad exposure metric as the predictor and the mental models scales as the outcome variables.
- Next, we'll fit two more regressions per two lines analysis: one left of the break point, one on the right. All of these additional regressions will have mental models as the outcome and feature a subset of the following as independent variables: (a) exposure to influencer VPN ads, (b) tech advice frequency, (c) iuipc-8 scores, (d) measured adoption of VPNs, (e) average spacing in between ads. To determine the exact subset, we'll do exhaustive model selection based on AIC, BIC, and adjusted R^2 with exposure to influencer VPN ads selected in each model.
- We will conclude that there is a relationship between exposure to vpn ads and mental models if alpha is .05 or lower for the exposure term in these regressions.

Exposure to is influencer vpn ads for participant i is defined as:
exposure_i = log((total_vpn_ad_words_i + 1) / mean_total_of_words_in_vpn_ads)

Total number of vpn ad words is determined by running the transcripts of each youtube video in participants' YouTube histories through a repurposed BERT model.

Additional analysis:
We will explore the relationship between exposure to VPN ads and brand familiarity with four ordinal logistic regressions, one each for the top four most frequently appearing companies in our final dataset. Each regression will feature X brand familiarity (measured with a likert) as the outcome variable and exposure to X brand influencer VPN ads as the independent. We'll additionally do model selection (same as described in the main analysis plan) to select from additional variables: tech advice frequency, iuipc-8 scores, measured adoption of VPNs, average spacing in between ads, and exposure to all influencer VPN ads except brand X.

For all regression analysis planned as a part of this work, ordinal independent variables will be bucketed into two buckets to increase overall power.

We'll run a Mann-Whitney U test to show the (possible) IUIPC-8 score difference between participants who finish the entire study and those who wish not to continue after the screener.

6) Describe exactly how outliers will be defined and handled, and your precise rule(s) for excluding observations.
- Low quality responses on multiple open ended questions.
- Participants who fail the attention check questions.
- Participants who we suspect have given us bad data (e.g., no or suspiciously little YouTube history).
- Participants who we were unable to find VPN ads for but self-report to have seen at least 10 influencer vpn ads in the past year.

7) How many observations will be collected or what will determine sample size?
No need to justify decision, but be precise about exactly how the number will be determined.
We're aiming to collect a minimum of 200 valid responses but will recruit as many participants as possible until we hit our maximum budget for recruitment. The current estimate based on preliminary data collection compensation rates is 210-225 participants.

8) Anything else you would like to pre-register?
(e.g., secondary analyses, variables collected for exploratory purposes, unusual analyses planned?)
We've collected a preliminary set of 36 complete responses. We used this set to inform the analysis plan described here. These responses will not be included in the final reported results. Thus, the sample sizes in 7 exclude these 36.

We additionally collected 196 partial responses to test our mental models questionnaires for internal consistency (cronbach's alpha) and sufficient variance. These participants did not complete the entire study and will not be included in our final analysis.

We're planning to run some analysis to see if any particular demographic groups are more likely to see ads or adopt VPNs but this analysis will remain exploratory.