Research Methods - Lecture 2: Introduction to research methods in psychology Flashcards
Population
All the scores, events, etc. that we’re interested in
-> NOT the individuals whom measurement/s are coming from
Sample
A representative subgroup of the population
Why is a sample taken?
We usually can’t measure the entire population we’re interested in, so we take a sample and infer general results about the population from the sample
2 ways sampling can go wrong
Sampling error and sampling bias
Sampling error (2)
- Results of repeated samples from the same population will always differ -> occurs by chance -> anytime take samples from populations, you are at the mercy of chance probabilities and may end up picking ppl who aren’t representative of the population
- Unavoidable
How can you minimise sampling error?
Using bigger samples
Sampling bias
Sample misrepresents population in a systematic way -> unrepresentative sample
How can you avoid sampling bias?
Random sampling - so every member of the population has an equal chance of being picked for the sample
What is the consequence of serious sampling bias?
It invalidates the research
What is the problem with random sampling in psychology?
Random sampling is very unlikely because psychological research focuses on behaviour of all humans and it’s very unlikely to get a representative sample therefore samples usually get volunteers - usually uni students (have time, live close by - more convenient for them to partake)
-> have to be cautious about whether you can infer anything from the sample in regards to the behaviour of the population overall
Does sampling error or bias relate to validity?
Sampling bias
Does sampling error or bias relate to reliability?
Sampling error
2 general kinds of research design
Observational and experimental
Observational design
- Measure two DVs and look for a relationship between them
* There is no IV, because nothing is manipulated - e.g., is self-esteem related to intelligence?
What is another name for observational design and why?
• Sometimes called a correlational design, because a relationship between variables is a correlation
When are observational designs used?
When the research/experiment conflicts with ethical or practical reasons
Basic weakness of observational design
Just because two DVs are correlated, we can’t conclude that one affects the other
CORRELATION DOES NOT IMPLY CAUSATION
-> Don’t know whether one variable A causes B or vice versa
-> Both variables could be affected by another third variable - Third variable problem
Experimental design
Manipulate IV and look to see effect on DV
Can imply causation (because controlling IV) - e.g., is self-esteem related to results of a fake IQ test?
-> Know direction of cause and effect
Which designs are more powerful?
Experimental designs are more powerful than observational, so we should use them when it’s ethical and practical to do so
When evaluating research that you read about what should you do?
Always consider this issue – has causation really been proved?
Fundamental principle of research design
We want to eliminate all explanations of our results except one - That is, if we conduct an experiment, and see a change in behaviour (our DV), we want to be sure that it was caused by our IV
What is a confounding variable?
- Alternative explanations of the results are called confounds or confounding variables
- A confounding variable is anything, other than our IV, that might have produced the change in DV that we saw
- They are bad, and we need to eliminate them if want our research to be worthwhile
General approaches to eliminating confounds - Standardisation
- Hold the confounding variable constant (standardisation)
* especially good for environmental or external confounds - as can manipulate these
General approaches to eliminating confounds - Randomise…
Randomize the confound
• especially good for subject-based or internal confounds - things you can’t change
General approaches to eliminating confounds - 3rd variable
• Decide that they’re more interesting and study them instead
You should always think about research that you design and that you read in these terms:
• What possible confounds were there?
• Have they been controlled?
• If not, don’t trust the research
-> If you can poke holes in research - repeat the experiment with adjustments to make up for the mistakes
Within-subject design (4)
Each subject is exposed to all levels of the IV (all experimental conditions)
• e.g., does alcohol affect short-term memory?
• Each subject tries to learn a list of words in both “sober” and “drunk” conditions
• There is an effect if individual subjects behave differently in the two conditions
Advantage of within-subject design
• Good for controlling subject confounds
Disadvantages of within-subject design
• But there could be environmental confounds – such as?
-> In alcohol experiment one way to avoid - keep list of words same
• Especially note the risk of an order effect - performance of second condition not due to IV being manipulated but just because it’s individual exposed to it second
-> Taking repeated measures - have to take at different time points - things change with time
2 ways of dealing with order effect
Randomise order - e.g. split half of the sample and each half starts off with opposite level of IV
Between-groups design
What happens when you measure the same individual twice?
Anytime you measure from the same individual twice, you do it over time and things change
Between-group design
- Each subject is only exposed to one level of the IV (one experimental condition)
- e.g., separate groups of subject in “sober” and “drunk” conditions
- Good for controlling environmental confounds
How can you tell if there is a difference between the two groups in a between-group design?
• There is a difference if the average behaviour of the two groups is different
How do you separate the groups in a between-group design?
Take random sample for each experimental condition
Everything between the two groups is exactly the same except the IV
Problem with between-group design and how to fix it
- But could be subject confounds – maybe the people in the “sober” condition just had better memories anyway
- Can partly prevent this by randomly allocating subjects to groups
Other names for between-group design
- Between subjects
- Independent samples -> taking many samples
- Independent groups -> each group only exposed to one IV
Other name for within-subjects design
• Repeated measures -> taking multiple measures
From results what can you not infer from between-group studies?
Can’t infer behaviour of individuals from either group because that study has not been done
e.g. can’t infer behaviour about alcohol group in a sober condition
What does within-subject design do that between-group design doesn’t?
Gives built-in test of generality
Matched-pairs design
- Give a pretest on memory ability before the experiment
- Find pairs of subjects with similar memory ability
- Randomly allocate one member of each pair to each condition, “sober” and “drunk”
How can you tell if there is an effect in a matched-pair design?
• There is an effect if the “sober” member of each pair consistently scores better than their “drunk” partner
Benefits of matched-pairs design
• Keeps environmental confounds constant, and (nearly) keeps subject confounds constant too
- > Enviro because each person is only exposed to one IV
- > Confounds in terms of what is being tested due to pairings of similar abilities
Con about matched-pairs design
• This is a very good design, but quite labour-intensive, so it isn’t used as often as it should be