Chapter 7 Lab 7: t-test (Independent Sample)

I think he [Gosset] was really the big influence in statistics… he asked the questions and Pearson and Fisher put them into statistical language, and then Neyman came to work with the mathematics. But I think most of it came from Gosset. —F. N. David

This lab is modified and extended from Open Stats Labs. Thanks to Open Stats Labs (Dr. Kevin P. McIntyre) for their fantastic work.

7.1 Do you come across as smarter when people read what you say or hear what you say?

7.1.1 STUDY DESCRIPTION

Imagine you were a job candidate trying to pitch your skills to a potential employer. Would you be more likely to get the job after giving a short speech describing your skills, or after writing a short speech and having a potential employer read those words? That was the question raised by Schroeder and Epley (2015).The authors predicted that a person’s speech (i.e., vocal tone, cadence, and pitch) communicates information about their intellect better than their written words (even if they are the same words as in the speech).

To examine this possibility, the authors randomly assigned 39 professional recruiters for Fortune 500 companies to one of two conditions. In the audio condition, participants listened to audio recordings of a job candidate’s spoken job pitch. In the transcript condition, participants read a transcription of the job candidate’s pitch. After hearing or reading the pitch, the participants rated the job candidates on three dimensions: intelligence, competence, and thoughtfulness. These ratings were then averaged to create a single measure of the job candidate’s intellect, with higher scores indicating the recruiters rated the candidates as higher in intellect. The participants also rated their overall impression of the job candidate (a composite of two items measuring positive and negative impressions). Finally, the participants indicated how likely they would be to recommend hiring the job candidate (0 - not at all likely, 10 - extremely likely).

What happened? Did the recruiters think job applicants were smarter when they read the transcripts, or when the heard the applicants speak? We have the data, we can find out.

7.2 Lab skills learned

  1. Conduct independent samples t-tests
  2. Discuss the results and implications

7.3 Important Stuff

7.4 JAMOVI

In this lab, we will use jamovi to perform an independent-samples t-test

7.4.1 Experiment Background

Schroeder and Epley (2015) conducted an experiment to determine whether a person’s speech (i.e., vocal tone, cadence, and pitch) communicates information about their intellect better than their written words (even if they are the same words as in the speech).

To conduct this study, the authors randomly assigned 39 professional recruiters for Fortune 500 companies to one of two conditions. In the audio condition, participants listened to audio recordings of a job candidate’s spoken job pitch. In the transcript condition, participants read a transcription of the job candidate’s pitch. After hearing or reading the pitch, the participants rated the job candidates on three dimensions: intelligence, competence, and thoughtfulness. These ratings were then averaged to create a single measure of the job candidate’s intellect, with higher scores indicating the recruiters rated the candidates as higher in intellect. The participants also rated their overall impression of the job candidate (a composite of two items measuring positive and negative impressions). Finally, the participants indicated how likely they would be to recommend hiring the job candidate (0 - not at all likely, 10 - extremely likely).

So, what happened? Did the recruiters think job applicants were smarter when they read the transcripts, or when the heard the applicants speak? We have the data, we can find out.

7.4.2 Performing an independent-samples t-test

First, let’s open the relevant data file; Here is the link. For our analysis, we will focus on only one of the three measures mentioned above: intellect. We want to know if perceived intellect is different in the audio condition (where recruiters listened to a job pitch) than in the transcript condition (where recruiters read a transcript of a job pitch).

Lucky for us, the condition variable is called CONDITION. Let’s take a look. There 0s and 1s for each condition (audio vs. transcript). But which one is which? This isn’t clear from the data, and it isn’t clear from the paper, or from the repository. We have to do some guess work. I went ahead and computed the means for the Intellect_rating between each condition, and then compared those to the graph in the paper for E4. It looks like 1 = audio condition, and 0 = transcript condition.

Let’s use words instead of 0s and 1s to refer to our experimental conditions. To do this, we will need to change the values of 0 and 1, to the words transcript and audio, and then run the independent samples t-test.To do this in jamovi, follow these steps, and you will see the rght side of the screen update with the chosen analyses and statistics:

We would summarize our results as follows: > Recruiter ratings for audio job pitches are 1.98 points higher than ratings for transcripts of pitches. This difference is significant, t(37)= 3.53, p<.05.

7.4.3 Practice Problems

  1. Use the same data file from this lab’s tutorial to test whether intelligence ratings were different between genders (use alpha = .05). Who was rated as more intelligent, males or females? Report your result in proper statistical reporting format.