EBM

research stats 1

types of data - categorical

• binary (disease/no disease, dead/survived)

• categorical, nominal (ethnicity, blood group)

• categorical, ordinal (alcohol consumption - no/low/moderate/high)

types of data - numeric

• count (no of visits to gp/y, integer valuesonly)

• continuous (BP/weight/height, measured on scale)

P value - probability of observed association to happen by chance

• small p (<0.05 - significant) probs not by chance, reject H0

• large p (>0.05 - not significant) probs by chance dont reject

risk ratio

• risk in exposed (a/a+b) divided by risk in unexposed (c/c+d)

• value below 1 → risk is lower in those exposed

odds ratio

• probability something happens divided by probability it doesnt

• ratio → odds in exposed (a/b) divided by odds in unexposed (c/d)

ratio interpretation

• 1 → outcome as likely to happen in one group as other

• <1 → outcome less likely in exposed group

• >1 → outcome more likely in exposed group