Independant/between T-test

Question 1

Normality checks

Answer 1

Is histogram symmetrical ?
Check central tendency- mean and median approx the same.
Kolmogorov-Smirnov.
Kurtosis/Skewness statistic should be less that 2 X std error.
If significantly different, exclude outliers; use non parametric test.

Question 2

Formal notation

Answer 2

Ignore paired samples correlation table.
Report t, df, and p (half for one tailed).
Relate to hypothesis.
Confidence interval.
Effect size: mean1-mean2/mean std dev. (0.2=small; 0.5=medium; 0.8=large).

Question 3

Standard error

Answer 3

Se= std dev(of sample)/square root N(in sample).

See how many se’s difference mean is along the curve: diff mean/se.

Question 4

Checks for similar variance

Answer 4

Homogeneity of variance.
Required when between subjects design and sizes of groups are very different.
Variance in two groups needs to be similar to calculate population parameters.

Question 5

Levene’s test

Answer 5

If sig value > .05 then variance between two groups is not significantly different. This is good.
If less, they’re different and equal variance not assumed.

Question 6

Reporting p values

Answer 6

Obtained from sig (two tailed) column.
If one tailed- halve sig value.
Always report p value with = sign.
Except when p

Question 7

Summary of results

Answer 7

Summary of data (no raw data).
Present graphs or tables.
Verbal description of any interesting results.
Justify use of test used.
Useful to report results.
Therefore a parametric t-test was used to analyse results.

Question 8

Restrictions for parametric tests

Answer 8

Data must be at least interval.
Sample should form or be from a normal distribution.
Between ppts- two data sets should have similar variance.