Selection of methods
Methods include: COQQIE
- Case studies
- Quasi Experiments
QUASI experiments is an experiment that does not fufill the requirements of a lab experiment. The IV is usually a fixed trait.
Eval. of Quasi Experiments
- Investigates real life differences and the effects they have.
- High ecological validity.
- Less control
- Less replicable
- Less establishment of cause and effect
SAMPLING STRATEGIES: Random
Researchers often have a target population they want to study, but this often too big to completely survey. Therefore, they must choose a sample.
Every person in a target population has an equal chance of being selected, and this must be done in an unbiased way. Methods include:
- Random number tables: They contain strings of numbers where each number has the equal chance of being chosen. They can be found in a statistics book, or generated by a computer.
- Computer selection: The computer generates a long list of random numbers with no relationship. Each name is given a number, and the random number generator programme is used to pick the sample.
- Manual selection: Put the sames (or assigned number) on a slip of paper, and into a container. The container is shaken, and then a blind selector must choose the participants in the sample.
SAMPLING: issues with random
- A truly random sample may still not be representative.
By chance, the randomization may choose similar participants. This is why research must be replicated to ensure this technue did not lead to unforeseen bias.
- Practical Limitations:
A very large target population would difficult to collect names, information, and consent from all participants. This takes time and effort.
- Refusal to participate
Participants selected may not want to take part, and will drop out. This can bias the sample also.
Takes a sample of people who are available at the time, and is NON RANDOM.
- Easy and quick
- Adequate for certain psychological processes e.g. memory.
- Has to be used for some natural experiments as the IV is not controlled.
- Biased sample; people in that area probably share something in common, and therefore, cannot be generalised to the whole population.
A Self selected sample is a technique that consists of particpants responding to a request by a researcher. This could be an advertisement. Non-random method.
- Quick and easy to use
- Reach a wide variety of people.
- Biased- people who volunteer for samples may not be representative. They may be more obedient/outgoing/ interested in psychology.
Sample Bias and generalisability
This is when a sample does not represent the target population.
- Enthusiastic participants- only those who want to be studied will remain, and so those who do not agree, are lost, creating a biased sample. The results will only be generalisable to people who agree to be studied.
- Choice of target population: Some dominant groups may be studied more than others e.g. cultural, gender and socio-economic bias.
- SMALL SAMPLES: are more likely to be biased.
Reliability: Researcher reliability
This is relevant where there is more than one researcher. Therefore, their ways of conducting and gathering data from the experiment may be different, resulting in poor reliability.
SPECIFIC TO OBSERVATIONS: Inter rater reliability it looks at the degree to which two raters agree on what they are seeing. If they record the same data exactly, this is high inter rater reliability.
- Design of the study should be controlled, precise and stardardised so that the instructions are easy to read and understand.
- TRAINING- the researchers should be trained to reduce variability in their behaviour.
The consistency in the measure and test of the independent variable within the study itself.
Also, if variables aren't controlled within the experiment, this can be called internally unreliable.
internal reliability: split half method
assess using the SPLIT HALF method:
- It assesses that all parts of the test contribute equally to what is being investigated.
- It compares the results of one half of the test with another half.
- If both sides produce similar results, there is high internal reliability.
- Any questions/observations that do not correlate with the other half can be removed, increasing reliability of the study.
PROS- quick and easy
CONS: can only be used for large questionnaires and observations that only measure one construct.
Reliability: External + assessment
This refers to the consistency of the results from one time to another and in one location and another.
Basically just repeat the study a few weeks/ months later. This checks that the same results would be obtained.
- ensures results are temporally reliable
- Takes a long time
- Expensive to re-do tests
- Time between studies must be long enough so participants don't remember answers (order effects) but short enough for not too many variables to change within the person, and so their views genuinely change. This would make the experiment seem unreliable.
How accurately a test or measurement measures what it says it measures. Does the IV cause the change in the DV, or was there another factor?
Can be lowered by
- researcher bias
- order effects
- demand characteristics
- individual differences
- extraneous variables
Validity: assessing internal validity x3
On the surface, is the study an accurate reflection of what it is trying to measure.
The scores from a new test/study are correlated against a old, accurate test are correlated. A positive correlation shows that the new test has high concurrent validity.
Does the measurement here give results that can be used to predict future events accurately.
Validity: How to improve x4
- Use highly controlled conditions to minimise extraneous variables and establish cause and effect.
- Double blind trials- neither the researcher or the participants know the hypothesis, therefore reducing researcher bias and demand characteristics.
- Order effects can be minimised by randomizing the order in which participants experience different conditons. Therefore, practice will not cause a bias towards one condition.
- Matched pairs- This controls individual differences by reducing the differences between the particpants in each condition.
Validity: External and assessment
This looks at how the findings of research can be generalised across people, places and time.
If a study lacks internal validity it automatically lacks ecological validity.
Assess external validity by:
- Population validity
The findings can be generalised across all cultures, socio-economic status and other variations within the human race.
- Temporal validity
The findings are as valid 50 years after they have been published, or were they subjected to bias from the time.
- Ecological validity
The findings can be generalised to other settings and environments. This can be either mundane validity (everyday settings) or experimental realism (where the participants believe the situation.)
Validity: What makes a externally valid study? x4
- Large sample size
- Representative sample
- Many repeats in different places/times.
- Recent- higher temporal validity.
R- right to withdraw
I- informed consent
P-protection from harm
LOOK AT AS CARDS
- Aim and Hypothesis-- null or experimental?
- Method x4
- Results x4
- mean and range
- statistical test
- chosen alpha level
- practical apps