Oxford English Dictionary: A scientific procedure undertaken to make a discovery, test a hypothesis or demonstrate a known fact.
Experimental vs. Non-Experimental Research
Experimental Research:-investigates causality
-manipulate IV to determine if it has an effect on the DV
-hold other variables constant
Non-experimental Research:
-observe and measure only descriptive
-investigates correlation/associations
True vs. Quasi Experiments
-True experiments randomly allocate participants to conditions.\
-Randomization: Each participant has an equal chance of being allocated to any condition.
-Why randomize? To reduce the chance of other variables (confounding oor nuisance variables) changing the outcome. Randomisation worksby spreading any potentially confounding variables evenly accross conditions. It minimizes systematic differences to isolate the causal effects of the treatment on the dependent variable.
-If it is not possible to radomise then a quasi-experiemental approach is used. Quasi-experiments do not use randomization. Also called 'natural experiments'.
-Types of Quasi-experiment:
1. One group pre-post test design: only treatment group, often used in audits to evaluate clinical services but outcomes may be due to other factors.
2. Non-equivalent control group design: A control group is used but randomisation is not. This means conrols are systematically different from the treatment group
3. Interrupted time series design: Do not alys have a control group, good for studying naturally occuring chronological data.
Systematica approaches to control for confounding:
1. Blocking/Randomised Block Design
>Used when a variable might have a particular influence on the outcome.
>Arrage the participants into groups (blocks) according to that variable.
>Randomized the participants within those groups to the treatment group.
>E.g. Gender
2. Matching/Matched-subjects/matched case-control design
>A statistical technique where every participant is compared with another participant who is matched according to a relevant confounding variable.
>E.g. Age, IQ
3. Counterbalancing
>In a within-subjects design the outcome may be affected by the testing order. Counterbalancing can be used to compensate.
>Latin Squares counterbalancing method is used with more than one condition/treatment. Testing is arranged so each condition occupies each rank in the testing order equally often.
没有评论:
发表评论