RESEARCH METHODS FLASHCARDS 1-3

Cause and effect relationship

an event that makes another event occur

Independent variable

what we change/manipulate

Dependent variable

what we measure

Operationalisation

making variables detailed and specific so they can be measured

Aim

general statement that the researcher intends to investigate, developed from research

Hypothesis

statement made at start of study (directional or non-directional)

Directional hypothesis

difference/relationship between two conditions is made clear (which DIRECTION)

Non-directional hypothesis

no previous research to suggest a direction the experiment will go in, 'there will be a difference' (two tailed)

Correlational hypothesis

relationships between two co-variables (can be directional or non-directional) predict positive or negative correlation

Null hypothesis

states there will be no significant difference between the two groups

Extraneous variables

any variables that could interfere with DV identified where possible at start of a study & measures put in place to minimise influence (age of participants, lighting, etc)

Confounding variable

variable from outside the study and may impact the result (personality of participants, time of day, etc)

Types of extraneous variables

situational (environment), participant (individual differences), investigator effects (influence from researcher), demand characteristics (deliberate change of behaviour)

Demand characteristics

how participants change their behaviour in regards to the study and info they pick up from it - may try to work out the aim of the stufy and act accordingly to how they think will support it (please-U effect) or how to cause issues (screw-U effect)

Investigator effects

behaviour of investigator may influence ppts and skew results, includes things such as body language and smiling, can influence those in the study to act differently from how they may have originally, unwanted influence

Randomisation

use of chance to reduce researcher's influence on design of investigation; can include randomly generating the order of words for a memory test or the order of conditions could be randomised
Key aspect is counter balancing - attempts to control order effects; half ppts take part in condition A then B and half B then A (REMEMBER ABBA)

Lab experiments

highly controlled environments, researcher manipulates IV and records effect on DV (Milgram obedience study)

Strengths of lab experiments

- High control over extraneous variable so you can conclude that the IV has had an affect on DV

- Replication is possible due to high level of control so results can be checked for reliability

Limitations of lab experiments

- Participants aware they're being tested which risks demand characteristics being observed

- Artificial environment so lacks generalisability, low mundane realism (doesn't reflect real life)

Field experiments

IV manipulated in everyday setting to see effect on DV

Strengths of field experiments

- Higher mundane realism because natural environment

- Behaviour is likely to be more authentic (less likely to be demand characteristics)

Limitations of field experiments

- Less control over extraneous variables

- Difficult to replicate (low reliability)

- Can be time consuming and costly

- ethical issues if participants unaware they're being studied

Natural experiments

Researcher takes advantage of pre-existing IV

'Natural' as IV would have changed even if researcher hadn't have studied it

Strengths of natural experiments

- Provides opportunities for research that may otherwise have not have occurred due to practical/ethical issues

- High external validity because includes study of real life

Limitations of natural experiments

- Naturally occurring events are rare so can't be generalised

- Pt's can't be randomly allocated so may be confounding variables

Quasi experiments

No control over allocation of pt's to conditions or the IV

IV could be based on existing difference between people, no one has manipulated the variable it just exists

Strengths of quasi experiments

- IV is a naturally occurring difference between people meaning changes in the DV may have more realism than if the IV was artificially created

- Pt's are likely to be aware they're being studied making consent easier to gain and so ethical issues

Limitations of quasi experiments

- Can only be used when naturally occurring differences between people can be easily identified (difficult to set up)

- Task used to further data for the DV may be unrealistic meaning the data has little mundane realism

- No random allocation leading to confounding variables

Independent groups/measures

involves different groups doing each condition (e.g. condition 1 consists of pt's A & B - condition 2 consists of pt's C & D)

Evaluation of independent groups/measures

Strengths: - Pt's are less likely to guess the aim
- Easy to set up and not time consuming

Limitations: - Individual differences may influence results

Repeated measures

Involves all participants experiences both (all) conditions, done to prevent order effects

Evaluation of repeated measures

Strengths: - By using all the same pt's in all conditions, there are no individual differences to act as confounding variables

Limitations: - By doing the experiment more than once in different conditions, the participants may be affected by order affects

Counterbalancing

order efefcts can be overcome by this. ppts split into 2 smaller goups, half ppts do condition A then condition B while the other half do condition B followed by condition A (ABBA)
This means effects of doing one condition after another in a repeated measures design with be counteracted

standardisation

try to ensure all participants experience research process in same way

random allocation

independent groups design - reduces potential of bias in groups

randomisation

researchers may randomise parts of the procedure to remove any bias by making it all 'due to chance' e.g. ppts may do the conditions in a random order by having a computer decide what order to present material to them

pilot studies

small scale trial run of experiment, allows you to check that design or method wil work, useful as practice to check for timings + instructions and to identify any issues
Changes can be made to ensure that the study runs smoothly and fewer errors are likely to occur in the procedure

matched pairs

different groups in each condition but the groups are matched on key factors

Random sampling

all members of target population have equal chance of being picked , randomly chosen through computer programme or picking names out of a hat

evaluation of random sampling

strengths - free from researcher bias as cannot choose people to 'help' their hypothesis
Limitations - difficult and time consuming because have to obtain info on every person in T.P., unlikely to reflect number of each member in society (unrepresentative), people may refuse to take part leading to a need for another randomiser

systematic sampling

every nth number of ppts is chosen to take part, sample frame is established (e.g. alphabetical order) and system is applied

evaluation of systematic sampling

strengths - avoids bias as researcher isn't directly choosing ppts, more representative than random sampling as more unlikely to end up with unbalanced sample
Limitations - still may not reflect correct quantities of each group in sample realistically, time consuming

Stratified sampling

target population broken down into it's key demographic components and ppts are selected from each strata according to it's relative size in the population

evaluation of stratified sampling

strengths - avoids researcher bias because calculated based on realistic results from target or general population, sample is representative because designed to accurately reflect composition of poulation so generalisation of findings is able to occur
Limitations - identified strata cannot reflect each person's different beliefs and feelings so complete representation is not possible, extrememly time consuming

Opportunity sample

anyone willing and available to participate is selected as difficult to obtain representative sample of T.P

Evaluation of opportunity sampling

strengths - convenient, saves time and money
limitations - creates bias as sample is unrepresentative as population drawn from neraby (very specific) cannot be generalised to rest of population, researcher bias as complete control over who partcipates

Volunteer sampling

ppts select selves to be part of sample. can be collected via adverts or a common room notice board, for example

evaluation of volunteer sampling

strengths - easy to do and requires less time + minimal input from researcher as ppts come to them
limitations - volunteer bias as may attract a certain 'profile' or person affecting generalisation of findings and leading to possible demand characteristics

What is reliability?

extent to which research is consistent, if research is repeated in same way with same findings results are said to be reliable

How can we improve reliability?

increasing objectivity will increase consistency over time and between people. such operationalised variables are essential.
standardising procedures with ppts is another way to ensure reliability internally + externally as ppts have consistent experience and study can be replicated to test reliability of results

What is validity?

extent to which the research actually measures and tests what it claims to

What is face validity?

whether a test, scale or measures appears on the 'face of it' to measure what it's supposed to measure

How can we improve external validity?

improved by developing realistic tests and using natural settings for studies rather than artificial tests in highly controlled conditions

What is internal validity?

extent to which the results of the study are due to the tested variable rather than extraneous or confounding variables

What is external validity?

extent to which results of study can be generalised to other people (population validity), other settings (ecological validity), and across time (temporal validity)

measures of central tendency

mean, median, mode

evaluation of mean

S: most sensitive measure as includes all reuslts making it more representative
L: also means it's easily distorted by extreme values, cannot be used on nominal data, harder to calculate than other measures

evaluation of median

S: unaffected by extremes, easier to calculate than the mean
L: less representative than mean as it doesn't use all scores

evaluation of mode

S: easy to calculate, unaffected by extreme values
L: data is often multi-modal so meaningless, doesn't use all scores

measures of dispersion

range and standard deviation

evaluation of the range

S: shows overall spread of whole dataset, easy to calculate
L: may not be representative of data if there's extreme values at the top and/or bottom of the dataset

Standard deviation

tells you the spread of data around the mean and allows you to see relationships between scores

Evaluation of Standard Deviation

S: provides representative measure of data spread, provides useful information about how individual scores relate to each other and to the mean, gives a measure of reliability of data as small standard deviations mean there was little variation in scores
L: harder to calculate than the range

correlation

illustrates strength ofdirection of an association between two or more co-variables
they're plotted on a scattergram where one co-variable forms the x
axis and the other forms the y axis

evaluation of correlations

S: useful for preliminary research in order to provide precise and quantifiable measures of 2 related variables, can suggest ideas for further research, quick and economical.
L: cannot show cause and effect, lack of experimental manipulation and control so extraneous variables are a variable, can be misued or misinterpreted

normal distributions

a symmetrical, bell-shaped curve meaning the mean, median and mode are all the same value and meet perfectly in the middle, most people located within the centre area with few people at either ends, tails never reach zero as extreme results are always a possiblity

right/positive skew

mean has the highest value then the median then mode as all different values

left/negative skew

mode has the highest value, then median, then mean so all values different

bar graph

consist of categorical data for comparison, categories go on x axis (bottom) and requency of occurrence on the y axis (side)

scattergrams

should be titled, one variable is plotted on the x axis, other on y axis, each axis should be labelled, for each pair of scores a dot/cross is placed on the graph where 2 scores meet, display correlational data

data tables

contain descriptive statistics such as measures of central tendency and dispersions. for some data, it may be limited to percentage calculations
all collumns and rows need to be lablled and have an appropriate title

histograms

these are used when the x axis is showing continuous data (bars touching) while the y axis shows the frequency of occurence of that data

Pie chart

circular statistical graphic divided into sections to illustrate numerical proportion, arc length of each sections is proportional to the quanity it represents

pie chart calculation

(requency/total) * 360

frequency polygon

show info from frequency table, first have to find the midpoints of each class, mid point of a class is the point in middle of the class e.g. a class "10-19" the midpoint is 14.5
Plot midpoints and join them up with straight lines

Nominal data

represented in categories, discrete as one item can only appear in a category

Ordinal data

data that has order, usually includes rating scales, doesn't have equal or universal intervals between each measurement

interval data

based on numerical scales with equal intervals between each measurement

statistical testing

provides way of determing whether hypotheses should be accepted or rejected, tells us whether differences/relationships between variables are statistically significant or occurred by chance

Significance level for sign test

always 0.05 unless otherwise specified

how results are significant with sign test

calculated value has to be equal to or lower than critical value

statistical test for nominal data + independent groups

chi square

statistical test for nominal data and repeated measures

sign test

statistical test for ordinal data and independent groups

mann-whitney

statistical test for ordinal data and repeated measures

wilcoxon

statistical test for ordinal data and correlation (relationship)

spearman's rho

statistical test for interval data and independent groups

unrelated t-test

statistical test for interval data and repeated measures

related t-test

statistical test for interval data and correlation (relationship)

Pearson's R

evaluation of nominal data

S: most simple, easy
L: only limited calculations can be made using it, superficial, measures of central tendency are innapropriate

evaluation of ordinal data

S: supports a more sophisticated analysis, enabling comparisons using measures of central tendency
L: do not have standardised interval scale so cannot guage options (like interval data)

How can you evaluate interval data?

best type of data for analysis as it supports all types of calculations and is suited to the most powerful inferential tests

When will you be asked to design a study?

In Paper 2, you're often asked to do this. It can be: an experiment, correlation or observation and can be for up to 12 marks

How would you answer a design a study question?

- Only write what the bullet points ask for; quite literally set up your work like this: "BULLET 1:", "BULLET 2:".. etc
- Make sure that your study is practical and replicable (e.g. use of standardised procedures, specificity, don't EVER say anything unethical, realistic samples)

Explain what is meant by probability

probability (p) is the likelihood that something will happen. It is expressed as a number between 0 and 1, where 0 means an event will definitely NOT happen and 1 means it definitely WILL.

Give an example/examples of probability

- the probability of getting tails when you toss a coin:
P = 1/2 = 0.5
- probability of getting a 6 when you throw a 6-sided dice: P = 1/6 = 0.16666... or 0.17

What are significance levels?

they're assigned to establish the probability of results being due to chance. If this is low, then we can reject our null hypo and accept our experimental/alternative hypo

In Psychology, what is the significance level we are willing to accept? Why?

1 in 20 likelihood (or p is less than or equal to 0.05)
This means we can be 95% confident that there is a true difference (experiment) or a relationship (correlation) between variables

Where might psychologists need to use stricter levels of significance?

In cases that have more dangerous implications such as in drug testing or in replications of studies where the aim is to verify original findings

Type 1 error

- wrongfully accept experimental hypothesis
- Believe there is a difference or a relationship when there isn't
- this is also known as a false positive or an error of optimism