Factors Affecting the Choice of Statistical Test
Which statistical tests do I need to know about?
- The Mann-Whitney test (non-parametric)
- The Wilcoxon test (non-parametric)
- Chi-squared test (non-parametric)
- Spearmans’ rho (non-parametric)
- Unrelated t-test (parametric)
- Related t-test (parametric)
- Pearson’s r (parametric)
The difference between parametric and non-parametric tests
- Parametric tests assume a normal distribution; non-parametric tests do not
- Parametric tests use interval data: non-parametric tests may use nominal or ordinal data
- Parametric tests assume homogeneity of variance; non-parametric tests do not
- Parametric tests are more powerful than non-parametric tests
What factors determine the choice of statistical test?
- Is it a test of difference i.e. a lab experiment or a test of association i.e. a correlation?
- Is the design independent measures, repeated measures or matched pairs (tests of difference only)?
- Is the data nominal, ordinal or interval?
- Nominal data is data arranged in categories; ordinal data is data which can be ranked without each value being equal in measurement; interval data is data where the intervals between each value are equal in measurement
Godden & Baddeley’s (1975) research on context-dependent forgetting using divers is an example of a test of difference: a ‘deeply’ interesting study…
TESTS OF DIFFERENCE |
TESTS OF DIFFERENCE |
TESTS OF ASSOCIATION |
|
RELATED DESIGN |
UNRELATED DESIGN |
||
NOMINAL DATA |
Sign Test |
Chi-squared |
N/A |
ORDINAL DATA |
Wilcoxon test |
Mann-Whitney test |
Spearman’s rho |
INTERVAL DATA |
Related t-test |
Unrelated t-test |
Pearson’s r |
Worked example
LEVEL: 4 MARKS
Bella has conducted research using elite athletes as her sample. She measured the body temperature of her participants who had either just run 10K or had rested for 30 minutes. Bella wants to carry out a parametric test on this data.
Explain whether or not Bella will be able to carry out a parametric test on this data. Justify your answer. [4]
AO2 = 4 marks
For full marks the answer should state that Bella should be able to carry out a parametric test as her data is interval (temperature measurements have distinct and equal intervals between each measurement); she can expect to see a normal distribution of data (because all of the participants are elite athletes they are likely to experience similar body temperatures during exercise or when resting; plus body temperature is a relatively stable variable). She can also expect homogeneity of variance (the standard deviations per condition are likely to be similar due to the nature of the sample – all elite athletes who are likely to have similar levels of health and fitness).