RMA: WEEK 4 Flashcards
Within-subjects design
- AKA repeated measure designs.
- All participants receive all levels of the independent variable. (e.g: all ppts experience different doses of a drug)
Between-subjects design
- AKA independent groups
- Different groups of participants receive different levels of an independent variable.
- e.g: Experimental and control group when testing the effectiveness of a drug.
Advantages of between-subjects design
- No order effects.
- Some experiments can only be between-subjects.
- Naïve participants > no learning in between so ppt wont improve significantly in second condition
Disadvantages of between subjects design
- Lots of participants. > expensive to gather for their time
- Characteristics might differ between groups.
e. g., how well they study generally, with or without music. (individual differences + participant variables) > don’t know if performance is due to IV or individual diff
How to counteract problems with between subject design
- Participants randomly assigned to groups > random allocation
- Match participants based on characteristics that may impact DV then assign to different conditions (some nuisance v cannot be controlled or some v matched however)
Advantages of within-subject design
- Fewer participants required. > cheaper
- Reduced individual differences.
e. g., confounding variables from each participant (essay writing ability) > Participants are used as ‘their own controls’. (matched to themselves)
Disadvantages of within-subject design
- Carryover effects. > Effect of one carries over to next session thus improves performance (e.g., ‘benefits’ from silent study affect following ‘with music’ behaviour.)
- Order effects > could get tired and put in less effort
How to counteract problems in within-subject design
- order of conditions could be randomised to avoid carryover effects
- conditions could be counterbalanced (A>B then B>A)
Counterbalancing
Half the participants do the conditions in one order than the other half do the opposite order (eg: 1/2 do A-B, 1/2 do B-A) there will still be order eff, results will just be more representative and general as an average
-Controls for order effects in repeated measures
Counterbalancing + latin square design
- if there are many conditions, there will be a higher amount of possible orders > inefficient to manually create this many orders so latin square resolves this
Latin square design
- Doesn’t use all possibilities available but has X amount of rows and columns where each condition occurs an equal amount of times at different points
- Carryover effects are still possible here > counterbalancing + latin square des doesn’t eliminate carryover effects but reduces them
- To find out how many possible combinations there are, you should multiply each condition by each other so if you want to completely counterbalance 5 conditions, you do 5x4x3x2x1=120 > 120 possible orders
Quasi-experiment
- One (or more) of independent variables are selected.
i. e., not manipulated. > Experiment where the IV is natural and re-occuring (eg age or gender) and impacts the DV - e.g: effect of education on memory skills (whether you have a degree or not cannot be manipulated)
- Participants not randomly assigned to the levels of the independent variable (education). > E.g., whether they have a degree or not.
Advantages of quasi-experiment
- Examination of variables that would be unethical to manipulate. > e.g: studying kids who have been institutionalised cannot be manipulated (cannot send a child to a care home), it is something which has already happened so it is ethical to study
Disadvantages of quasi-experiment
- Leads to more confounding variables that cannot be removed. > E.g., people with better memories more likely to go to university?
- No strong conclusions about cause and effect possible. > many other variables could impact the person’s memory (e.g heritability)
How to counteract problems with quasi-experiments
- Matching participants + correlating
- If treatment study: tests before and after treatment.
(participant are their own control)