STATS1- testing for differences between 2 sample means

tests of difference + between subject design

• parametric = independent t-test

• non-parametric = mann-whitney

tests of difference + within subject design

• parametric = paired t-test

• non-parametric = wilcoxon

tests of relationship/association + continuous variables

• parametric = pearson’s r

• non-parametric = spearman’s rho

test of relationship/association + categorical variables

chi-squared

cohen’s d formula (population & sample)

small, medium, large cohen d values

• small= 0.2

• medium= 0.5

• large= 0.8

what does a small cohen’s d value suggest

most chance of overlap so small effect size

what does the null hypothesis suggest for testing differences between 2 means

mean A & mean B are equal

what does the research hypothesis suggest for testing differences between 2 means

• mean A > mean B

• mean A < mean B

• mean A & mean B aren’t equal

what does mean A < mean B mean

difference in sample means is less than 0 (mean A - mean B < 0)

what does mean A > mean B mean

difference in sample means is more than 0 (mean A - mean B > 0)

6 steps for paired t-test

1. hypothesis

2. data collection

3. calculate difference in paired scores between conditions (post-pre)

   • mean change = mean post - mean pre

4. reformulate hypothesis (mean change =/< 0)

5. calculate t-score & convert to critical value

6. reject or fail to reject null hypothesis

what does it mean is t-score > critical value

p < 0.05

z-score for independent t-test

t-score for independent t-test

bottom = eseA² + eseB²

how to work out V for independent t-test

total sample size - 2

4 steps for independent t-test

1. hypothesis

2. collect data: mean, sd, ese, n

3. calculate t-score and assume null hypothesis is true (pop mean diff = 0)

   • convert t-score to a critical value

4. reject or fail to reject null hypothesis