RM Lecture 3 - Independent Samples T-Tests/Between Subject T-test

T-tests

how likely are we to obtain the observed difference between conditions if the null hypothesis is true? Is the variance between conditions larger than the variance within conditions?

T statistic

t=variance explained by our IV/unexplained variance

comparing means

is the diff between two means significant?

normality testing

a common assumption for statistical tests is that our data are normally distributed

Shapiro-Wilk test

bell-shaped curve means there’s a high probability of getting the results by chance (we want a high probability); Shapiro wilk statistic, if the p value is bigger than 0.05 then the data is normally distributed and passes test of normality

one sample t-test

compare data from your experiment with one single number, eg did participants score sig higher than the national average; one dependent variable, no manipulation as such, everyone does same test

requirements for one sample t-test

data are independent, continuous and normally distributed

reporting the results

data handling (what you’ve done to your data), descriptive statistics, say which test you’ve done, mention relevant assumptions, report test statistic, df and p-value

non-parametric alternative for one-sample t-test

one sample wilcoxon signed-rank test

independent samples t-test requirements

data are independent, interval/ratio (continuous), n of at least 12, data normally distributed, homogeneity of variance

homogeneity of variance

variance in condition 1 similar to variance in condition 2; look at descriptive statistics, visualize the data, run a homogeneity of variance test (Levene’s test)

Q-Q (quantile-quantile) plots

data should fall roughly on a straight line, ordered data, sectioned into ‘quantiles’ and plotted

Levene’s test

is the variance in condition 1 sig different to the variance in condition 2?, we want the answer to be no (p>0.05, no sig diff)

non-parametric alternative to independent samples t-test

Mann-Whitney U Test if data fails homogeneity of variance only you don’t need to run a separate test