Understanding Experimental Analyses and Statistical Tests
2.4 Experimental Analyses
Overview of Experimental Analyses
Objective of Experimental Analysis
To understand how the results of experiments are analyzed.
Unlike correlations, true experiments provide solid evidence about cause and effect.
True Experiments
True experiments rely on random assignment to groups:
Ensures each experimental condition is equal before the experiment begins.
The independent variable is the only possible difference in outcomes between competing groups.
Example Experiment: Multitasking Ability
Scenario of Experiment:
The hypothesis is that spending more time on social media during a lecture (independent variable) influences exam scores (dependent variable).
Design:
Recruit 100 students.
Randomly assign:
50 students browse social media for 30 minutes during the lecture.
50 students browse social media for 10 minutes during the lecture.
Each student is tested on the lecture content at the end of the class.
Data Analysis
Data Analysis Process:
Upon gathering data, statistical analysis is necessary.
The analysis compares two groups using the t test statistic.
The t Test
Definition:
The t test is a statistical analysis that compares averages and ranges of two groups to determine if they differ significantly.
Historical Context:
Developed by William Sealy Gossett, a chemist and statistician, while working as the head brewer at the Guinness Brewery in Dublin, Ireland.
Created to monitor quality between morning and afternoon batches of Guinness without quality sampling all kegs.
Key Components of the t Test:
Random Samples: Must avoid bias through random selection.
Sample Size: The number of samples must be sufficient to represent all beer casks produced during a run.
Publication Under Anonymous Pen Name:
Gossett published his findings under the pseudonym Student, which is why the t test is sometimes referred to as Student's t test.
Alternative Naming:
Some suggest it would be more appropriate to refer to it as Gossett's t test.
Application to Multitasking Experiment
In the multitasking experiment:
The scores of participants are displayed in visual form (e.g., Figure 2.5).
Group comparisons:
Green Group (10 minutes): Performed better than the Blue Group (30 minutes).
Comparing Three or More Groups: Analysis of Variance (ANOVA)
Context for Use:
If there are more than two groups to compare, such as in an extended version of the multitasking experiment.
Example Setup:
Five groups:
Group 1: No social media (control group).
Group 2: 10 minutes on social media.
Group 3: 20 minutes on social media.
Group 4: 30 minutes on social media.
Group 5: Entire class duration on social media.
Analysis of Variance (ANOVA):
ANOVA compares three or more groups, determining if one group significantly differs from the others.
If differences are found, further investigation is needed to identify which groups are significant outliers.