Module 1

Categorical and Quantitive

Divides the cases into groups/ categories

measure a numerical quantity for each cases

Nominal and Ordinal

Discrete and Continuous

information collected from whole (finite) population

Information collected from a subset of the \n population

***Often use the approach to sampling (and data collection) to distinguish between options...

to make inferences back about Population Parameters allowing for RANDOMNESS

by measuring relationships between variables at a point-in-time (correlations)

RANDOM SELECTION! without it, we get biased

Treatment variable and outcome variable

EXPLANATORY VARIABLES

attributes we can measure that explain variation in the Outcome

e.g: The Treatment variable is just a ‘special case’ of an \n explanatory variable

CONFOUNDING VARIABLES

attributes we cannot measure that explain variation in the Outcome

e.g: With COVID-19 this might be attitudes to social distancing, approach to hygiene and isolation, or your job situation

• Can have SEVERE BIAS due to background effects

• Correlations are not Causation

• guiding the development of hypotheses for later experiments

  • We look for evidence of a response in a group getting the treatment BEFORE implementing a large expensive double-blind randomised controlled experiment

• Observational studies are useful for reinforcing the results of double-blind randomised controlled experiments

the participant does not know if they are \n Treatment or Control

neither the administrator of the experiment or the participant knows if they are Treatment or Control