Rationale

In framing theory, some outcome – like an attitude, judgment or action – depends upon exposure to a cue embedded in a media message. The cue prompts the message consumer to apply a particular interpretive frame to the message.

In the experiment, the ad is the message, and the body type of the actor shown in the ad is the cue. Framing theory suggests that the actor body shown in the “muscular guy” ad would encourage ad viewers to apply a “healthy food” frame to the ad, whereas the actor body type shown in the “rounder guy” ad would encourage viewers to apply an “unhealthy food” frame to the ad.

In both instances, a cue embedded in the ad would prompt the ad’s viewers to interpret the ad’s information in a particular way.

Hypothesis

The average calorie estimates from the two groups will differ significantly.

Variables & Method

The continuous variable named “estimate” measured each participant’s estimate of the number of calories in the sandwich shown in the ad. The categorical variable named “group” indicated whether each participant belonged to the group whose ad featured the slim actor or to the group whose ad featured the heavier actor. The analysis treated “estimate” as the dependent variable, and treated “group” as the independent variable. Given this combination of dependent and independent variable characteristics, an independent-samples t-test is the most suitable test for evaluating the hypothesis.

Results

First, the analysis examined descriptive statistics for the group distributions and tested the distributions for normality. Calorie estimates by the group that viewed the slim actor’s ad averaged 212 calories, with a standard deviation of 91.0. Meanwhile, calorie estimates by the group that viewed the heavier actor’s ad averaged 292 calories, with a standard deviation of 92.4.

Group distributions showed some possible outliers, especially among participants who viewed the heavier actor’s ad. But Shapiro-Wilks tests found no evidence of nonnormality in either distribution. These findings suggested a independent-samples t-test would be suitable for the analysis, as planned.

The test found the difference between the group averages to be statistically significant, t (77.97) = -3.92, p < .05. Thus, the results supported the hypothesis. Specifically the results suggest that calorie estimates by the group that viewed the slim actor’s ad averaged significantly lower than did calorie estimates by the group that viewed the heavier actor’s ad.

Code and output

# Installing required packages
if (!require("dplyr"))
  install.packages("dplyr")
if (!require("tidyverse"))
  install.packages("tidyverse")
library(dplyr)
library(ggplot2)
options(scipen = 999)

# Read the data
mydata <- read.csv("SandwichExperiment.csv") #Edit YOURFILENAME.csv

# Specify the DV and IV
mydata$DV <- mydata$estimate
mydata$IV <- mydata$group

# Graph the group distributions and averages
averages <- group_by(mydata, IV) %>%
  summarise(mean = mean(DV, na.rm = TRUE))
ggplot(mydata, aes(x = DV)) +
  geom_histogram() +
  facet_grid(IV ~ .) +
  geom_histogram(color = "black", fill = "#1f78b4") +
  geom_vline(data = averages, aes(xintercept = mean, ))

# Calculate and show the group counts, means, standard
# deviations, minimums, and maximums
group_by(mydata, IV) %>%
  summarise(
    count = n(),
    mean = mean(DV, na.rm = TRUE),
    sd = sd(DV, na.rm = TRUE),
    min = min(DV, na.rm = TRUE),
    max = max(DV, na.rm = TRUE)
  )
## # A tibble: 2 × 6
##   IV           count  mean    sd   min   max
##   <chr>        <int> <dbl> <dbl> <int> <int>
## 1 Muscular guy    40  212.  90.6    30   380
## 2 Rounder guy     40  292   92.4    40   450
# If you see evidence of non-normality in the distributions of one or both groups, use a Shapiro-Wilk test.
mydata %>%
  group_by(IV) %>%
  summarise(`W Statistic` = shapiro.test(DV)$statistic,
            `p-value` = shapiro.test(DV)$p.value)
## # A tibble: 2 × 3
##   IV           `W Statistic` `p-value`
##   <chr>                <dbl>     <dbl>
## 1 Muscular guy         0.976     0.531
## 2 Rounder guy          0.970     0.358
# A significant Shapiro-Wilk test for one or both groups means you should use a Wilcoxon signed rank test rather than a t-test.

# options(scipen = 999)
# wilcox.test(mydata$DV ~ mydata$IV)

# If both groups are normally distributed, though, you may run a t-test.

options(scipen = 999)
t.test(mydata$DV ~ mydata$IV,
       var.equal = FALSE)
## 
##  Welch Two Sample t-test
## 
## data:  mydata$DV by mydata$IV
## t = -3.9223, df = 77.967, p-value = 0.0001875
## alternative hypothesis: true difference in means between group Muscular guy and group Rounder guy is not equal to 0
## 95 percent confidence interval:
##  -120.98249  -39.51751
## sample estimates:
## mean in group Muscular guy  mean in group Rounder guy 
##                     211.75                     292.00