data_simmonscredibility: SimmonsCredibility - Ch14 - from Simmons and Nelson (2020)
In rcalinjageman/esci: Estimation Statistics with Confidence Intervals

data_simmonscredibility

R Documentation

SimmonsCredibility - Ch14 - from Simmons and Nelson (2020)

Description

You're excited! Your company has developed a wonderful new weight-loss program, and now it's your job to develop the ad campaign. Should you choose a BeforeAfter pair of pictures, as in Figure 14.1, top panel? Or might a Progressive sequence of pictures of the same person, as in the bottom panel, be more effective? Pause, think, and discuss. Which would you choose, and why? You might think that BeforeAfter is simpler and more dramatic. On the other hand, Progressive highlights the steady improvement that you claim the program will deliver. You're probably not surprised to learn that BeforeAfter is used often and has long been a favorite of the advertising industry, whereas Progressive is used only rarely. Luca Cian and colleagues (Cian et al., 2020) were curious to know the extent to which BeforeAfter is actually more effective, appealing, and credible than Progressive, or, indeed, whether Progressive might score more highly. They reported seven studies of various aspects of that question. I'll focus on their Study 2, in which they used three independent groups to compare all three conditions illustrated in Figure 14.1. The BeforeAfterInfo condition, in the middle panel, comprises three BeforeAfter pairs, thus providing extra information about the before and after endpoints. The researchers included this condition in case any advantage of Progressive might stem simply from having more images, rather than because it illustrates a clear progressive sequence. They randomly assigned 213 participants from MTurk to one of the three groups. Participants were asked to 'imagine that you have decided to lose some weight', then saw one of the three ads for a weight loss program called MRMDiets. They then answered the question 'How would you evaluate MRMDiets?' by choosing a 1-7 response on several scales, including Unlikeable-Likable, Ineffective-Effective, and Not credible-Credible. The researchers averaged six such scores to give an overall Credibility score, on the 1-7 scale, with 7 being the most credible. Simmons and Nelson (2020) were sufficiently intrigued to carry out two substantial very close replications. With the cooperation of the original researchers, they used the same materials and procedure. They used much larger groups and preregistered their research plan, including data analysis plan. I'll focus on their first replication, in which 761 participants from MTurk were randomized to the three groups.