2. Comparative: two groups Flashcards
List the statistic test you can use for differences between groups (one variable)
PARAMETRIC:
> One sample t-test
> Independent samples t-test
> Related (paired) t-test
NON-PARAMETRIC
> Mann-Whitney U-test
> Wilcoxon
What 3 values are typically reported with statistical tests? Why?
- Test value (X2, t, or U)
- Degrees of freedom
- P value
To ensure that our test are transparent, and that they can be checked by others.
What’s degrees of freedom?
It’s the number of values in the final calculation of a statistic that are free to vary.
In stats, d.f. can refer to the number of cases minus the number of means.
For within group designs, the test uses the mean difference between time 1 and time 2.
What does testing normality do?
This is basically testing how well the data fits the model of normal distribution.
If the results are not significant, then the data is normally distributed.
What’s the T-test?
> Parametric test
Comparative stats test
Measures the difference between two means.
Variations of the t-test
> One sample: a group average different from a known value.
> Independent/between subjects
> Paired/related: Within subjects
Independent T-test
Between subjects. Comparing the scores of two mutually exclusive groups of participants.
Variances must be similar in the groups.
> Need to check ‘Levene’s test for equality of variances’. This must NOT be significant. Otherwise, we can’t assume there is equal variance.
What is variance?
The square of the standard deviation.
Paired/related T-test
Within subjects. Comparing scores from the same participants at two different time points, or on two different skills.
> Is also used when one group only is measured on two different skills, or under two different conditions.
> Basic requirements are the same as for an independent T-test.
Non-parametric tests and conditions
Independent samples –> Mann-Whitney
Related samples –> Wilcoxon
They require data expressed on a continuous scale BUT the sample size must be small, and the data can be skewed.
Conditions of Mann Whitney U test
Use this test when: > Design is independent groups/between groups > Parametric assumptions are not met > There are ONLY 2 groups > Data is continuous
(SEE NOTES ON HOW TO CALCULATE!!)
Conditions of the Wilcoxon matched pairs Test
Use this test when: > The design is matched groups / within groups > Parametric assumptions are not met > There are ONLY 2 sets of data > Data is continuous
Calculating Wilcoxon (W-test)
Rank the different scores, ignoring the direction of the difference.
Some differences will be positive and negative.
The W is the sum of ranks of the differences with the least occurring sign.
How do we report NON-PARAMETRIC TESTS
> U/w and p-value
> NO DF REPORTED (unlike t-test)
