Experiments

Control group exists

Intervention group exists

Both groups have statistically insignificant/negligible differences. Allocation of control and intervention is randomised.

Randomisation allows a causal relationship to be concluded from the data

X treatment, Y outcome, Z confounders

Z are a problem because they obfuscate the effect of X

Randomisation breaks link between X and Z to allow Y to be measured solely through effects of X

Randomisation also allows the control and treatment to be same on average

There exists a null hypothesis which can be disproven.

There can be a simple cross section experiment design with no followup. There can be one or two followups to an experiment to see the changes to the participants.

Data collection methods Nonresponse
 * Computer(interview, self-interview, telephone, surveys)
 * Mail
 * Telephone
 * Personal Interview
 * Types
 * Unit nonresponse - nonresponse for the entire survey
 * Item nonresponse - nonresponse for one question or a subset of questions
 * Sources
 * Refusal
 * Inability
 * Fatigue
 * Unavailability
 * Non-contact
 * Mistake/accident
 * Adjusting for nonresponse
 * Weighting
 * Imputing
 * Followup
 * Incentives
 * Substitution


 * Result: Nonresponse bias, inaccurate survey result and generalisation