Session 4 - Statistical testing Flashcards
Introduction to Statistical Testing
Purpose of statistical testing
To identify if any relationships between variables are statistically significant or not
Tests try to disprove the…
Null hypothesis
H0
Base assumption is H0 is true
Aim to reject in favour of alternative hypothesis HA, if test is significant
Null hypothesis
H0
Proposes there is no statistical significance
Claims the effect being studied does not exist
2 sided vs 1 sided test examples
Do children from larger families have better resilience?
2 sided
H0 - Average resilience scores are the same
HA - Average resilience scores are different
1 sided
H0 -Average resilience scores are equal or lower
HA Average resilience scores are higher
1 sided or 2 sided test?
Generally use 2 sided to allow for the possibility of results against our expectations
p-value interpretation
> 0.1 little or no evidence
0.05-0.1 weak evidence (not statistically significant, fail to reject H0)
0.01-0.05 Evidence
<0.01 Strong evidence
<0.001 Very strong evidence
Type I error (alpha)
Deciding against a true null hypothesis
(test statistically significant, but no true difference = H0 is true)
Type II error (beta)
Failure to detect a true alternative hypothesis
(In population HA is true, there is a difference, but test is not statistically significant)
Confidence Interval - the relationship is statistically significant if….
the interval does NOT include the value of the null hypothesis (difference=0)
What to include in write up of statistical testing
Descriptive Summary of variables including direction and magnitude of any observed relationships
Relevant test output
Comment on statistical significance - exact p-value and quote CI of any difference
Comment on test assumptions of the test that was performed
Inferential statistics
Make judgements about a population of interest using data from a sample of participants
