Types of test Flashcards
How are 2 independent groups typically compared
- Randomised study (invention vs control)
- Cohort studies (exposed v non exposed)
What is assumed when doing an unpaired t-test?
- Variable is normally distributed
- Standard deviation is similar in two groups
- participants are independent between groups
Which hypothesis is tested in an unpaired t-test?
That there is no difference between the two groups (null) (the most boring one)
Interpretation of P value
P < 0.001 (strong evidence against null hypothesis)
P < 0.05 (moderate)
P > 0.1 (little)
P = 1 (none)
What is a paired test design?
Participants are paired up (eg age, gender) then put in separate groups so the groups are separate but paired
Measurements are taken before and after intervention and then compared
What is analysis variance?
ANOVA test
a method for hypothesis testing - proves a global p-value comparing been across all groups
What is assumed when hypothesis test mean across 3 or more independent groups
- Similar SD
- Independent participants
- Normally distributed / non skewed variable
when data is paired how does analysis of variance vary?
it is repeated
Assumptions when comparing mean across 3 or more paired groups?
difference scores normally distributed or not too skewed
SD of differences scores should be similar for any combination of groups
when is a post hoc pairwise comparison used?
when p is < 0.05 and you want to compare groups to each other
when using non-parametric methods how should groups be summarised?
using medians & interquartile ranges
when should non-parametric methods be used?
when the assumptions for parametric methods are not met
what test is used when comparing a quantitative variable between two independent groups?
Mann Whitney test
what test is used when comparing a quantitative variable between 3 or more independent groups?
Kruskal-Wallis test
what test is used when comparing a quantitative variable between 2 paired groups?
Wilcoxon signed ranks test
what test is used when comparing a quantitative variable between 3 or more paired groups?
Friedman test
What are advantages of non-parametric methods
always valid for quantitative data
provide similar P-values when parametric assumptions are not met
disadvantages of non-parametric methods?
do not make direct inferences about a parameter
no confidence intervals
based on analysis of ranks not actual scores
less powerful the parametric methods when assumptions are met
How can normal distribution be checked for
plot histogram or box and whisker plots for each group
How can similar SD be checked
SD x SD = variance
variance for one group should be no more than 4 times another group
Standard deviation definition
A quantity showing how much the members of a group differ from the mean value for the group
Standard error definition
a measure of how accurately a sample’s distribution represents the true distribution of the population