Table 1 Results of Linear Mixed-Effects Model for Studies 1, 2a, and 2b

From: Warning people that they are being microtargeted fails to eliminate persuasive advantage

 

Dependent variable:

 

Persuasiveness

 

(Study 1)

(Study 2a)

(Study 2b)

ad_typepopup

-0.026

-0.033

-0.063

 

(-0.097, 0.044)

(-0.106, 0.039)

(-0.126, -0.001)

 

p = 0.466

p = 0.370

p = 0.046

ad_typenon_targeted

-0.215

-0.149

-0.171

 

(-0.369, -0.061)

(-0.287, -0.011)

(-0.328, -0.014)

 

p = 0.007

p = 0.035

p = 0.033

matching_score_centered

0.036

-0.162

-0.091

 

(-0.049, 0.120)

(-0.319, -0.004)

(-0.212, 0.031)

 

p = 0.407

p = 0.045

p = 0.143

ad_typepopup:matching_score_centered

0.052

0.163

-0.002

 

(-0.060, 0.164)

(-0.030, 0.356)

(-0.148, 0.143)

 

p = 0.365

p = 0.098

p = 0.977

ad_typenon_targeted:matching_score_centered

-0.210

-0.048

-0.139

 

(-0.354, -0.067)

(-0.268, 0.172)

(-0.310, 0.032)

 

p = 0.005

p = .668

p = 0.112

Constant

2.690

2.748

2.832

 

(2.570, 2.809)

(2.624, 2.871)

(2.704, 2.959)

 

p < 0.001

p < 0.001

p < 0.001

Observations

59,940

38,880

60,210

Log Likelihood

-86,243.700

-57,251.780

-89,003.000

Akaike Inf. Crit.

172,525.400

114,541.600

178,044.000

Bayesian Inf. Crit.

172,696.400

114,704.400

178,215.100

  1. Results of the generalized linear mixed-effects model used in Study 1 (left), Study 2a (center), and Study 2b (right), with self-reported persuasiveness as dependent variable. Estimates are provided together with their 95% confidence intervals. In all models, the observations (i.e., the persuasiveness ratings) are reported individually rather than being averaged within subjects. To address the fact that each participant rated multiple ads on various scale items, and that the same ads appeared under different conditions, random slopes for ad type were included for both participants and ads. Additional variables considered in the model include ad type (with ‘targeted-no popup’ used as reference level) and the matching score, which has been mean-centered within each ad type and represents the scaled difference between the participant’s openness score and that of the ad. Both ad type and the matching score are included in the formula as a two-way interaction.