Stats cram Flashcards
Purpose of cross sectional study
Prevalence of exposure and disease at specific point in time
Purpose of ecological study
Assess link between exposure and outcome
Used if outcome of study is rare
Purpose of cohort study
Studies incidence rate and whether exposure is associated with outcome of interest
Used for large study populations
May be prospective or retrospective
Purpose of RCT
Determines possible effect of specific intervention on population
Minimises bias
Variants of RCT
- Parallel group design - participants remain in assigned groups for entire duration
- Crossover design - participants switch groups during study
- Factorial design - experimental study design in which multiple interventions are studied simultaneously
- Cluster design - experimental study design in which unit of randomisation is a group rather than individual participant
Types of errors
Type 1 error - null hypothesis rejected when it is actually true, or alternative hypothesis is accepted (false positive error)
Type 2 error - null hypothesis is accepted when it is actually false (false negative error)
Statistical power
1 - probability of type 2 error
P-value
Compared to significance level
Chance of type 1 error
Purpose of T-test
Calculates difference between means of two samples
Unpaired - two different group samples at same time
Paired - same group samples at two different times
Purpose of ANOVA
Extension of t-test
Calculates difference between more than 2 groups
Purpose of Spearman correlation coefficient
Calculates relationship between two variables according to rank
Purpose of Mann-Whitney U test
Compares ordinal, interval or ratio scales
Calculates whether two independently chosen samples originate from sample population
Purpose of Wilcoxon test
Rank sum - Compares means between groups of different sizes
Signed rank - compares means between pairs of scores that can be matched
Purpose of Kruskal-Wallis H test
Compares multiple groups by testing null hypothesis
Purpose of binomial test
Examines whether observed frequency of event with binary outcomes is statistically probable or not