t-tests

Question 1

t-test

Answer 1

• Inferential statistics
• We can use a t-test to determine if our experimental manipulation (IV) has had a significant effort on our DV
  
  • Compares the difference between two sample means to differences we would expect to find by chance
  • If observed difference is much larger than expected, they are samples from different populations
  • The expected difference is based on the SE
• Types of t-test
  
  • Independent measures t-test → between-groups design
  • Dependent measures t-test → within-subject design
  • One-sample t-test → when we have one group only and we want to compare these participants’ scores to a specific score
• One-tailed significance: when you have a one-tailed hypothesis, halve the significance value p

Question 2

Paremetric test assumptions e.g. t-test

Answer 2

1. We have independent scores
2. Data are measured at the interval/ratio level
3. Data are normally distributed
   
   • For independent t-test, data for each group should be normally distributed
   • For dependent t-test, differences between conditions should be normally distributed
4. Just for independent t-test: Assumption of homogeneity of variance should be met

Question 3

Homogeneity of variance

Answer 3

Levene’s test needs to be non-significant (p>.05) for assumption to be met

Question 4

Reporting descriptive statistics

Answer 4

Question 5

Reporting t-statistic

Answer 5

Question 6

Effect sizes

Answer 6

• Standardised measure of the size of an effect
• Objective measure
• Cohen’s d: the difference between the two means divided by the SD
  
  • There are two different SDs, so use pooled SD values
• Report with t-statistic
• Cohen’s d guidelines
  
  • 0.20 = small effect
  • 0.50 = medium effect
  • 0.80 = large effect