Introduction to experimental design
The psychology experiment • Predict the causal effect of one thing on another
• Keep everything constant other than the affecting thing
• Vary the affecting thing systematically
• Measure changes in the affected thing
• Assess statistically whether or not the predicted effect has occurred
Within participants
same participants in each condition controls for individual differences introduces order effects and carry-over
effects
Overview Between participants different participants in each condition no order effects Individual differences need to match or randomise allocation
of participants
Mixed design
Mixed design some conditions have different
participants, some have the same
Example- between subjects design Comparing the number of errors made
entering into a computer spreadsheet for a sample of people listening to loud popular music with the number of errors made by a different control sample listening to white noise.
Two different people are compared
Examplewithin-subjects design Studying the number of keyboard errors
made by a group of 20 secretaries, comparing the number of errors when music is played to when music is not played.
Performance of one group of people is compared in two different circumstances
Why distinguish We need to choose the appropriate
statistical test: Between – unrelated or uncorrelated t –
test Within - related or correlated t-test
Why laboratory research ?
Practicalities: equipment/apparatus to bulky, security, expensive
Experimental control: keeping all factors the same E.g. light, temp, noise, arrangement of equipment
These are extraneous or environmental factors
True or randomised experiment Experimental manipulation: manipulated
variable = independent variable e.g. Alcohol Alcohol increases the number of mistakes The level/amount of alcohol = IV Amount given to each subject is constant for
each condition
Condition one = 8ml and condition two = 16ml Lower quantity of alcohol = control condition Higher quantity of alcohol = experimental
condition
Control group: View nonviolent film
Randomly assign into control and
experimental groups
Full population of interest
Experimental group: exposed to independent variable: view violent film
Checks on experimental manipulation Experiment on memory and anger Researcher says pre-scripted offensive comments to people in the experimental group and nice things to the control group Possible problems:
View it as a joke, patronising Resolve the issue by either:
get subjects to complete of questionnaire on their mood after debriefing ask how they felt about the researchers questions Pilot
Standardisation of procedures Keeping things constantAlcohol and error experiment
Time of day Body weight of participants Time they ate Researchers behaviour Any others ?
Resolutions Tape recorded instructions Come into lab previous day
Randomisation Who goes in the experimental or control group What if the participant undergoes more than one condition
Toss of a coin more than two
Throw of a dice Write on cards, random number tables, computer number generation
Problems: runs of the same condition or number of participants in either condition is different Randomisation ensures that there is no systematic bias in the selection process of participants, although chance factors may lead to differences between the conditions.
Matching Ensuring equal numbers Matched block or block randomisation
First Ss of a pair is randomly assigned to control condition using the specified procedure, while other pair is assigned to remaining condition
We need to ensure that participants in the control and experimental condition are similar Matching on gender, age weight
Pre-test and post test sensitisation effects Without a pre-test
there is only a measure of people Performance after drinking
But, look at the pre-test – maybe due to randomisation people who generally made more mistakes were in the 8ml group
8ml
16ml
Number Of
errors
pre-test post-test
14
10
6
2
Cont… Having a pre-test helps us to determine whether randomisation worked It allows us to see whether or not there has been a change in performance between the pre- and post test Disadvantage
Alert the Ss to the purpose of the experiment Solutions Increase the length of intervals between the pre and post test We could test participants again after the post test
Within-subjects design Fatigue or boredom – number of mistakes
maybe more in the second than in the first condition
Practice effect – Ss become better at task
Carryover, asymmetrical transfer – the effect of an earlier condition affects the subsequent condition. Solution increase time between conditions, but the problem is sometimes they just don’t come back !!
statistical significance The key to determining if a treatment had an effect is to measure the
statistical significance. Statistical significance shows that the relationship between the variables is probably not due to mere chance and that a real relationship most likely exists between the two variables.
Statistical significance is often represented like this: p < .05
Cont…. A p-value of less than .05 indicates that
the possibility that the results are due merely to chance is less than 5%. Occasionally, smaller p-values are seen such as p < .01. There are a number of different means of measuring statistical significance. The type of statistical test used depends largely upon the type of research design that was used.
AndrogynyAndrogyny
Androgyny
Today we accept a lot more diversity (e.g. Hayley Cropper off Corrie) and see gender as a continuum (i.e. scale) rather than two categories. So men are free to show their “feminine side” and women are free to show their “masculine traits”. For example, • Beckman wears a skirt • Earrings for men • Women’s boxing • Girl Power So it has become a lot more difficult to say what us typically “male” or “female”, and people who are biologically one sex often possess qualities (and the behaviour) appropriate to the opposite sex.
AndrogynyAndrogyny Refers to the recognition that individuals possess qualities (or traits) which are characteristic of both masculinity and femininity (Bem, 1974) Davison (2000) - women that those who had androgynous characteristics scored highly in terms of their well-being, than women that were not androgynous. Gana (2001) found that highly androgynous husbands had a happier home life and participated more in the household tasks and in the bringing up of the children than did husbands with rigid traditional gender views.
Questionnaire Questionnaire Take 10 minutes to complete this questionnaire and score it. Do not identify yourselves on the questionnaire
Lets do your first psychological experiment !!!!
We will use these results for our seminar session next week, and create a discussion section ourselves during the seminar session. I will provide you with the introduction and methods sections.