Research Methods A Flashcards
Learn the content
Hypothesis
Derived from theories, they are testable predictions
Independent variable
A variable varied by the experimenter in order to examine the effects of the dependent variable (Is tested on)
Dependent variable
A variable liable to be influenced by the independent variable (what is measured in the experiment)
Which is independent and dependent variable?…
“Eating carrots improves eyesight”
Independent: Eating carrots
Dependent: Eyesight
Three problems with research
Can be bias
Can breach ethics
Confounding variables
Confounding variable
An extraneous variable that has interfered with the results of the experiment
Three ways to avoid bias in an experiment
Single blind study
Double blind study
Use a placebo for a group
Single blind study
The participants are kept in the dark about specific elements of the study
Double blind study
The participants and the researcher conducting the experiment are kept in the dark about specific elements of the study
What Clever Hans tells us about bias
That it is better to conduct a double blind study so the examiner can’t give unconscious physical clues as to the correct or preferred answer
Reactivity
When the knowledge that a participants is being observed or measured influences their behaviour
Three types of research methods
Non-experimental
Experimental
Quasi - experimental
Four types of non-experimental research
Observational
Case study
Survey
Correlational research
Difference between non-experimental and experimental research
Non-experimental research is descriptive whereas experimental is explanative and contains control factors
How are observational experiments carried out?
Mainly through categorization with as little disturbance as possible
Example of observational experiment
Eibl-Eibesfeldt’s cross-cultural eyebrow raising during greeting observations
Two problems of observable methods
and how they are solved
-Reliability of categorisation (due to subjectivity)
Solved by comparison with other researchers
Reactivity of subjects
Solved with observers undercover as participants
How are case studies carried out?
Observation of a single person or particular group, often with a unique quality
Three problems with case studies
Generalisations
Reproducibility
Lack of cause and effect understanding
A solution to the problems with case studies
Deviant case analysis: create a situation similar to the case study with a distinct difference to work out the cause effect relationship
Three types of surveys
Questionnaire
Interview
Diary study
Four problems with surveys
Reactivity
Validity of questionnaire
How to quantify
Participant’s memory
Benefits of a structured interview (three)
Easily quantified
Comparable across participants
All topics covered
Three costs of a structured interview
Rigid structure Not personally adaptable Surface information (can't probe deeper depending on participants answers)
Two benefits of an unstructured interview
More in-depth information
Personalised to participant
Two costs of an unstructured interview
Generalisability
Analysis can be time consuming (especially for big groups)
Purpose of correlational research
Determine the relationship between two variables without manipulation
Problems with correlational research (four)
Confounding variables (secondary causes etc)
Can often be unclear
Can be coincidental
Correlation is not proof of any causation
What is an experiment?
Manipulate the independent variable to test the effects on the dependent variable
Null hypothesis
The idea that there is no relationship, nothing happening in the study. Is always assumed while conducting the study
Nuisance variable
Additional factor that may effect the dependent variable (the results)
How to stop nuisance variables turning into confounding variables
Either turn it into a control variable or intentionally make it another independent variable
Control variable
Variables kept the same or otherwise made sure to not interfere with the dependent variable
How to solve nuisance variables across participants (E.g. sex, age etc.)
Either separate systematically or spread randomly
Why including multiple independent variables in one experiment rather than multiple experiments is better (three)
More efficient
More control over nuisance variables
Can see the relationships of independent variables
How to carry out a multiple independent variable study
Make sure to include groups for each possible combination of independent variables
How to tell on a graph if there is interaction between independent variables or not
If there is an interaction, the lines for each independent variable wont be parallel, if there isn’t an interaction, they will be parallel.
The more unparalleled the lines the stronger the interaction
Common example of multiple dependent variables in an experiment
Speed and accuracy (has to be trade off)
Advantages of experiments (two)
Stronger test of causality
Possibility of a variety of manipulated controls
Disadvantages of experiments (three)
Unnatural setting/task causes more reactivity
Some phenomena cannot be studied this way (E.g. social interaction)
Ethical limitations
What happens to a nuisance variable when it is not dealt with by the experimenter?
Becomes a confounding variable
A within subject design experiment
All participants receive all levels of the independent variable
A between subject design experiment
Different groups of participants receive different levels of the independent variable
Between subject design advantages (3) and disadvantages (2)
Adv: No order effect, essential for some experiments, naïve participation
Disadv: Lots of participants, characteristics between groups may differ (can be solved)
Within subject design advantages (2) and disadvantages (2)
Adv: Fewer participants, reduces individual differences
Disadv: Carryover effects (into next IV tested), order effects
How to counteract order effects in within subject design experiments
Randomise, or better: The Latin Square Design… make each order occur equally often (cant have too many variables though!)
Quasi experiment (and example)
When one (or more) independent variables are selected - not manipulated
Eg) Education relation to memory
IV: university degree YES/NO
DV: score in memory test
Advantage and disadvantage of Quasi experiment
Adv: Can examine otherwise unethical variables (as no manipulation)
Disadv: Possibility of confounding variable means no strong causal conclusions
Three types of sample
Random
Stratified
Quota
A random sample
Everybody in cohort has an equal chance of being selected
Why is random sampling particularly difficult
Always depends where you are to select people (choosing people ‘at random’ outside tennis court would be opportunity sampling)
Stratified sample
Random selection of each subgroup of the population/cohort
Quota sampling
Represents prechosen proportions
Four psycho-physiological measurements
Muscle activity
Eye movement
Blink rate
Brain imaging
Advantages (2) and disadvantages (3) of EEG scanner
Adv: excellent temporal resolution, relatively inexpensive
Disadv: Poor spatial resolution, surface activity of brain, artefacts (confounding variables in brain activity)
Advantages (2) and disadvantages (4) of FMRI scanner
Adv: excellent spatial resolution (2-3mm), Accesses all brain areas
Disadv: Poor temporal resolution (5sec lag), expensive, participants cannot move, those with claustrophobia may refuse
Milgram’s obedience study
Participants instructed by man in white coat to give an actor an electric shock
(thought unethical due to psychological damage to participants)
When can ethics be deprioritised
When the potential research could have a large positive impact on society
Why is fully informed consent of the participant not always possible? And how is this made up for?
Due to potential reactivity from the subject
Made up for with a full debrieft afterwards
Three rights of the participant
To leave at any time
To have their privacy valued
No sharing of their information without consent
What groups of people have different ethics applied to them?
Children, seriously ill people, prisoners (feel a pressure to cooperate for benefits)
By who is all national research ethically monitored by?
The British Psychological Society (BPS)
Why do researchers experiment on animals? (2)
They are models of humans
Generally viewed as less valuable than a human
Guidelines to animal research are imposed by… (2)
Animal Act 1986 The BPS (very strict)
Problems with measuring variables (2)
Subjectivity (eg, mood or intelligence)
Testability (eg, mood or anxiety levels)
Types of measurement (4)
Nominal scales
Ordinal scales
Interval scales
Ratio scales
Nominal scale and example
Numbers are labels, no relationship between the size and attribute measured (eg, bus numbers)
Ordinal scales and example
The order of the size of the number equals order of the size of the attribute measured (distance between scores vary) (eg, IQ score, medal table)
Interval scales and example
Equal interval on scale is the same as the equal interval in property measured (eg, degrees C)
(can have negatives)
Ratio scale
Equal interval on scale is the same as the equal interval in property measured, and 0 denotes an absolute absence (eg, time taken…)
Difference between interval and ratio scales
Interval can be negative, ratio cant
The mean
Sum of scores / number of scores
The median
Midpoint of sample
if n is even, is the mean of the middle two scores
When to use the mean and the median and mode
Mean: no or little anomalies
Median: Some anomalies, ordinal data
Mode: Nominal data
Mode
Most frequently occurring
Bimodal
Two modes in the data
Limitations of the mode (3)
Some data doesn’t have one
Some data is bimodal (or more)
Can be atypical (not typical of data)
Alpha value
Generally when the p value is 5% (0.05), it is the point at which researchers reject the null hypothesis
(there is a 5% chance of their hypothesis being incorrect - probably not down to chance!)
p value
Value generated for the probability of the results being due to chance
When do researchers change the alpha value
Will lower it when the consequences of the results are serious and raise it when they are trivial
Measures of spread (3)
Range
Interquartile range
Standard deviation
Range
Maximum value - Minimum value
Interquartile range
Measure of spread between the middle 50% of scores (Third quarter (Q3) - first quarter (Q1))
Standard deviation
Measure of variation around the mean (the higher the value the larger the spread)
How to calculate standard deviation
- Find the mean
- Find the deviation (how far each score is away from the mean)
- square all the deviation scores (makes all positive)
- divide by n
- square root (to reverse initial squaring)
The variance
standard deviation squared
Equation for standard deviation
sqr rt of: sum of (x - mean)^2 / N
Why use graphs (3)
Indicates data patterns
Helps decide how to use data
Illustrates findings to others
When are bar graphs good to use
Ordinal or nominal data
Advantages of histograms (2)
Area shows frequency of data
Clearly shows the mode and outliers
How to do a stem and leaf diagram
Stem represents the start of number, leaf is the end
What do box plots show (6)
Minimum value Q1 Median Q3 Maximum value Any outliers
How are outliers defined in box plots
1.5 x above or below Q3 and Q1 respectively
What do scatterplots show
The relationship between variables
What does a perfect correlation on a scatterplot mean (2)
Either fake data or the same thing is being measured in different ways (measuring height in cm and inches)
Purpose of correlation analysis (3)
Determine the nature, direction and strength between two variables
Two correlation coefficients
Pearson (r)
Spearman (r s)
Does changing the units of measurement effect correlation
No
What does a non linear correlation look like
Line of best fit is curved
Example of non linear correlation
Relationship of stress and resilience
How is the Pearson correlation coefficient calculated
Directly from the raw scores
When to use Pearson correlation coefficient (3)
Suitable for interval and ratio data
Little to no outliers (is highly affected)
Not for skewed data
How us the Spearman correlation coefficient calculated
Ranking of the raw scores
When to use Spearman correlation coefficient
Suitable for ordinal data
Can be used with outliers (marginally effected)
Suitable for skewed data
Density curve
Basically a histogram with a curve of best fit from each top point (shows distribution of population)
When are density curves good to use
With lots of participants (generalise the population)
What does positively and negatively skewed data look like
Lump on the left for positively skewed data and the right for negatively skewed data
What does the area under the distribution curve equal (2)
1 or 100%
Finding the median, upper quartile and lower quartile in distribution curve
Median splits area underneath in half
Quartiles split it in quarters either side (Q1 / Q3)
Normal distribution
A lot of naturally occurring data (height) is distributed symmetrically around a single central dendency
Symbol for: 1) sample and 2) population mean
3) Standard deviation, 4) population standard deviation
1) x bar
2) myou (u)
3) S
4) theta (o)
How are the location and shape of the normal distribution curves determined
Location by population mean (u) and shape by population standard deviation (o)
Where does the tail of the normal distribution curve meet the x axis
At infinity (or never)
What do statistical tests assume about the distribution of data
Normally distributed
If data is not going to be normally distributed, what kind of test is needed
Non-parametric tests
Why normal distribution curves are useful
Can compare different data sets (eg two exams)
How do you compare normal distribution curves
Translate them into standardised normal distribution curves and calculate the z-scores, then plotting them onto standardised curve
What do z scores show
The number of standard deviations that the score is away from the mean
z-score = …
x - mean / S
What does the mean and S equal in a standardised normal distribution curve
Mean = 0 S = 1
What do you do after plotting the z-score on the standardised normal distribution curve
Look up in table on on SPSS the decimal area to the left of the point’s vertical line from the x axis.
(minus from 1 to get the area on the right)
Type 1 error
Rejecting the null hypothesis when it should not be rejected
Why wouldn’t we always lower the alpha level to 0.01 to avoid type 1 errors?
Because of type 2 errors
Type 2 error
Failing to reject the null hypothesis when we should reject it
Generally, what is the probability of a type 1 error
5%, assuming the alpha value is 0.05
Generally, if the null hypothesis is accepted, whats the probability of a type 2 error
= p. E.g. if alpha = 0.05 and p = 0.08, the probability that we should have rejected the null hypothesis is 8%
??
Directional hypothesis (one tailed)
Specifies which way the results will go (should be based off prior research)
Non-directional hypothesis (two tailed)
Only predicts some form of difference / relationship
Does the null hypothesis change depending on if a directional or non-directional hypothesis is chosen
No it doesn’t, is always the general alternative to hypothesis (there will be no relationship…)
What can samples show about a population, and how is this accounted for
Can only show an inference, a margin of error often calculated to account for this