Research Methods P2 Flashcards
Hypothesis
-Clear, Precise and testable statement
Should be operationalised - how you will measure …
-states what variables to be investigated
-stated at start of the study
Directional Hypothesis
-One tailed
States direction of the difference
ex. More than
Done when there is previous similar research
There is a a positive difference between IV and the DV…
Non-Directional Hypothesis
-Two tailed
States a difference between two variables
The is a difference between IV & DV
Aim
A general expression of what the researcher tends to investigate
-The PURPOSE
Independant Variable
-Manipulated by researcher to investigate the effect on The DV
Dependant Variables
- What the researcher is measuring
- Effected by change in IV
Extraneous Variables
- nuisance variables -> harder to find result/ detect the effect
- Must attempt to control EV’s
Participant Variables
- Type of Extraneous Variable
- Features of the participant, individual differences -> e.g. a persons mood
Situational Variables
- type of EV
- Features of the situation, e.g. are the instructions standardised?
Confounding Variables
When SV and PV occur at the same time as IV, they become Confounding Variables
- Change systematically with the IV
- Must be controlled
Operationalisation
- what the variables are defined by
- How they will be measured
Demand Characteristics
refer to any cue given by the researcher or situation, that may reveal the aim of the study & change participant’s behaviour
- > ‘SCREW -U’ effect - underperform
- > ‘PLEASE-U’ effect - overperform
Investigator effects
Any effect of the investigators behaviour on the outcome of the research, the DV, or design decisions
->’Expectancy effects’ is where the investigator provides unconscious clues
Randomisation
the use of chance when designing investigations to control for the effects of bias
e.g. random allocation of participants to conditions
Standardisation
- using the exact same formalised procedures for all participants in a research study
- if not differences can become EV’S
Hypothesis writing tips
- is the IV and DV clear & measurable
- are the differences in direction between IV & DV mentioned
- Is it two tailed or one tailed
Experimental Design
different ways of organisation for testing participants (RIM’ed)
- Independant groups
- Repeated Measures
- Matched Pair Design
Independant groups
-One group does Condition A the other does condition B
-PPTS are randomly allocated to experimental groups
+ no order effects -> tested 1, can’t practise/get tired, will not guess the aim ->behaviour more ‘natural’ -higher realism
- participant variables, ppts in 2 groups are different -> acting as EV/CV -> reduce validity
- Less economical -> more expensive than if done once as more ppts needed and more time used recuiting -expensive
Repeated Measures
- All ppts take part in all conditions of an experiment
- Order of conditions should be counterbalanced -> avoid order effects
+ participant variables, the people in both conditions have the same charachteristics
+ Fewer participants -> more economical as saves time recruiting and money spent
- order effects, may do worse/better when doing similar task twice ->practise/fatigue effects
- Ppts guess aims -> reduce validity of the results as a change in behaviour
Matched Pair design
-2 groups of participants - related as paired on participant variables
+matched on variables relevant -> control ppt variables & enhances the validity
+no order effects -> only doing condition once -> no fatigue/practice effects
- matching isn’t perfect as time consuming
-need twice as many ppts than RM for same data -> time spent recruiting
Laboratory experiment
- controlled environment, EV & CV regulated
- ppts go to researcher
- IV is manipulated -> effect on EV is recorded
+EVS & CVs can be controlled -> effects are minimised -> cause & effect between IV and DV can be demonstrated -> high internal validity
-lacks generalisability -> artificial tasks & ppts may be aware that they are changing the study -> low external validity
Field Experiment
a natural setting, researcher goes to participants
-IV is manipulated & effect on DV is recorded
+more natural -> produce more authentic behaviour -> more generalisable
+ppt’s are unaware they are being studied -> no change in behaviour due to demand characteristics ->greater external validity
- more difficult to control CV/EVs -> changes in DV = CV/evs -> hard to establish cause & effect
Natural experiment
-IV is not manipulated, changes naturally -> it will change without an experimenter.
-IV varies
-DV is naturally occurring or may be devised by experimenter & measured in a field/lab
+only practical/ethical option, greater external validity ->involves real-world issues like stress & exams ->more relevant & valid
-natural event may be rare, ppt’s aren’t randomly allocated
Quasi experiment
IV based on pre-existing difference between ppl e.g. gender -> just exists
DV can be naturally occurring or devised by the experimenter and measured in field/lab
+high control, comparisons can be made between people
-ppt’s not randomly allocated, casual relationships not demonstrated
Population
the large group of people that a researcher is interested in studying
Sample
smaller group of larger population selected to take part in the study
Generalisation
sample should be representative of population so generalisations can be made
Bias
Most samples are biased in that certain groups may be under or over represented
Random sample
every person in target population has an equal chance of being selected
- by lottery method, all members of TP are given a number and placed in a hat/computer randomiser used
+potentially unbiased -> EVs/CVs are controlled -> enhances internal validity
-time consuming, lists are hard to get, ppt’s may refuse to take part
Systematic Sample
ppt’s are selected by using a ‘set’ pattern.
-every Nth person is selected from a list of the target population
+unbiased -> selected at random as it is an objective method
-time and effort -> could just use random
Stratified sample
-The sample represents proportions of people in certain subgroups within population
-subgroups are identified -> percentages of subgroups are reflected in the sample
+ representative method,
- stratification is not perfect
Opportunity sample
people who are most available -> easier to obtain
-ask ppl nearby to participate
+quick method -> cheaper
- invetiably biased -> unrepresentative of TP as it’s from one area e.g. street
Volunteer sampling
-uses participants which select themselves
-by advertising in media e.g. newspapers
+ppt’s are willing -> engage more
- volunteer bias -> share certain traits -> respond to cues & generalisation is limited
Ethical issues
- when a conflict exists between the rights of a participants and the aims of the research
- BPS code of conduct is a quasi-legal document to protect ppts
- Respect, Competence, Responsibility and Integrity
- Committees weigh up costs & benefits of the study to decide whether it proceeds
Informed consent
- ppt’s make a informed judgement about whether to take part
- may reveal the aims
- seek consent forms, appropriate seek parental consent
Types of consent
- Presumptive: ask a similar group and assume the ppt’s agree
- Prior general: agree to be deceived
- Retrospective: Get consent after the study
Deception
- issue is deliberately misleading or withholding information so consent is not informed, mild deception is okay
- should be debriefed & told the real aims, details that weren’t given during study, what their data will be used for and the right to withhold data
Protection from Harm
- issue is that participants should be at no more risk than they would be in everyday life
- given right to withdraw at any stage
- should be reassured of their behaviour & that it was normal during debriefing
- Researcher should provide counselling if distress in ppt’s
Privacy & confidentiality
- the issue is that we have the right to control information about ourselves -> invaded = confidentiality should be respected
- held personal details must be protected, researchers refer to ppt’s using numbers, initials or false names
- personal data can’t be shared with other researchers
Correlation
illustrates the strength & direction of an association between two co-variables
Scattergram
one co-variable is on x-axis
other on y-axis
Types of correlation
-Positive = co-variables increase together
-Negative = one co-variable increases as one decreases
-0 correlation = no relationship between variables
on scale from -1 to +1 if close to one the correlation is stronger -> more significant
if further away from one the correlation is weaker -> less significant
Experiment vs Correlation
experiment - Researcher manipulates IV and records effect on DV
correlation - no manipulation of variables, cause & effect no demonstrated
Correlation evalution bp
+Useful starting point for research -> provides a measure of how two variables are related -> suggest hypotheses
+ relatively economical ->no need to control variables & use secondary data -> less time consuming than experiements
-no cause & effect -> presented as casual as only show how 2 variables are related -> false conclusions
-intervening variables, have untested variables which are involved -> false conclusions
Observational techniques
- way of seeing or listening to what people do without having to task them
- Often used as a way of assesing the DV
+capture what people do -> ppl act differently from self-reports -> provide insight into behaviour
-risk of observer bias -> may interpret situations differently due to expectations -> bias can reduce by using 2 observers
Naturalistic -OT
Takes place where target behaviour will usually occur
+High external validity -> more generalisable
- Low control -> can be uncontrolled CVs/EVs -> more difficult to detect patterns
controlled -OT
-some control/manipulation of variables including control of CVs/EVs
+can be replicated -> easily repeated due to standardised procedures -> check findings
-low external validity -> can’t be applied
Covert -OT
-participants are unaware they are being studied
+ demand characteristics reduced -> behaviour is natural -> increases internal validity
-ethically questionable -> ppl may not want to be recorded -> participants right to privacy
Overt -OT
-participants are aware they are being studied
+ethically acceptable -> informed consent -> right to withdraw
-demand characteristics -> knowledge of being studied influences behaviour -> reduces internal validity
Participant observation
-researcher becomes part of group they are studying
+ greater insight -> R experiences situation as participants do -> enhances external validity
-loss of objectivity -> R may identify too strongly with group -> threatens objectivity & internal validity
Non-Participant observation
-researcher remains separate from the group they are studying
+ more objective -> researcher maintains objective distance, less chance of bias -> increase internal validity
-loss of insight -> reduce external validity
Behavioural categories - observational design
- the target behaviour to be observed should be broken up into a set of observational categories
- difficult to make clear & unambiguous -> categories must not overlap
- dustbin categories -> all forms of behaviour should be in the list
Event sampling -OD
-a target behaviour/event is recorded everytime it occurs
+ useful for infrequent behaviour that could be easily missed in time sampling
- complex behaviour oversimplified -> important details go unrecorded -> affect validity of findings
Time sampling -OD
-observations are made at regular intervals e.g. once every 15 seconds
+reduces number of observations -> data recorded at certain intervals -> observation is structured
-unrepresentative - not reflect whole behaviour
Questionnaires - self report techniques
a pre-set list of written questions to which a participant responds
+can be distributed to lots of people, straightforward to analyse
-responses may not the truthful -> social desirability bias, response bias -> may favour a kind of response
Interviews - self report techniques
face to face/online interaction between an interviewer and interviewee
types: Structured, Unstructured, semi-structured
Structured Interview
a list of predetermined questions asked in a fixed order
+easy to replicate -> reduces differences between interviewers
-interviewers can’t elaborate -> limit the richness of data collected
Unstructured interview
no set questions, general topics discussed
Interaction is free-flowing
Interviewee encourages to elaborate
+greater flexibility -> gains more insight into interviewee’s worldview -> can collect unexpected information.
-Increased risk of interviewer bias
Semi-structured interviews
A List of questions that have been worked out in advance
but interviewers ask further questions based on given answers
Designing questionnaires
-avoid jargon
-avoid double barrelled questions
-avoid leading questions
whether you will use closed or open questions?
Closed questions
-limited choice of responses
-produces quantitive data (numbers, scales) & qualitative data -> convert it
+Easier to analyse ->easier to draw conclusions
-responses are restricted, forces an unrepresentative answer -> reduces the validity of the findings.
Open questions
-respondent provides own answers expressed in words
-produces qualitative data (opinions)
+responses not
restricted -> have more external validity
-difficult to analyse -> forced to reduce data to statistics
Designing interviews
A standardised list of questions that needs to be covered -> reduce interviewer bias
- a quiet room will encourage the interviewee will open
- Begin with neutral questions -> ppt’s relaxed
- Remind interviewees that answers will be treated in confidence
Pilot studies
A small scale trial run of a research design using a small number of participants
-used to find out if certain things don’t work so you can correct them before spending time & money
Qualitative data
non-numerical data expressed in words
+richness of detail ->more meaningful leading to greater external validity
-Difficult to analyse -> leads to subjective interpretation & researcher bias.
Quantitative data
numerical data
+easier to analyse
-Narrower in meaning -> lower external validity
Primary data
-first hand data collected for the purpose of the investigation
+fits the job -> info is directly relevant to research aims
-requires time & effort -> secondary data can be accessed within minutes
Secondary data
-collected by someone other than the person who is conducting the research
+inexpensive -> requires minimal effort
-Quality may be poor ->information may be outdated -> challenges the validity
Meta-analysis
a type of secondary data, involves combining data from a large number of studies.
+increases the validity of conclusions -> increases the extent to which generalisations can be made
-publication bias ->therefore conclusions may lack validity
Measures of central tendency
mean
median
mode
Mean
an average
-add up all scores & divide by the number of scores
+sensitive measure -> includes all scores in data -> representative
-can be unrepresentative ->one extreme result can distort the mean
Median
the middle value
places scores in ascending order & select middle value
if two median values use the mean of both
+Less affected by extreme scores
-less sensitive than the mean
Mode
most common value, used in nominal data
+relevant to categorical data
-overly simple measure -> not useful when many modes
Measures of dispersion
range
standard deviation
Range
difference between highest and lowest score
+easy to calculate
-doesn’t account for the distribution of the scores
Standard deviation
Measure of the average speed around the mean
larger the standard deviation, the bigger the spread of data
+more precise than the range
-it may be misleading
tables
raw scores displayed in columns & rows
a summary paragraph beneath explains the findings & draws conclusions
Bar charts
categories are placed along the x-axis and frequency on the y-axis
the height of each column represents the frequency of that item
Histograms
bars touch each other
data is continuous
a true zero
Scatter grams
used for correlational analysis
each dot represents one pair of related data
illustrates strength & direction of correlation
Normal distributions
symmetrical, bell-shaped curve
most items are in the middle area of the curve
mean,median,mode occupy the same mid-point of the curve
Skewed distributions
distributions that lean to one side because most items are either at the lower or upper end of the distribution
Negative skew
-distribution is towards the right of the graph, resulting in a long tail on left
mode is highest peak, median in middle and mean is dragged left
Positive skew
- most of distribution is towards the left of the graph, resulting in a long tail on the right
- mode is highest point, median is middle, the mean is dragged to right
significance
-Significance, the difference between two sets of data is greater than what would occur by chanc
Probability
- about how likely it is that a certain event will happen if the null hypothesis were true
- accepted at 0.05
sign test
-analyse related designs with nominal data
-tests a difference
Subtract B from A -> produce the sign of difference
Add up - &+
Omit those who stay the same -> deduct from N
S value is total of the less frequent sign
if S is equal or less than critical value, S is significant
Peer review
- before publication, all aspects of the investigation are scrutinised by experts in the field
- objective
- peer should be unknown to researcher
Peer review aims
funding: allocate research funding
validation of the quality & relevance of the study
Improvements & amendments are suggested
peer review eval
+protects quality of published research -> minimises possibility of fraudulent research -> preserves reputation of psychology as a science
- anonymity can be used criticise rival research
- Publication bias
- Ground breaking research may be buried
the economy
- findings of psychological research can benefit our financial prosperity
e. g. Attachment research into the role of father, stressed role of father, promotes flexible working arrangements -> contribute to economy
e. g. The development of treatment for mental disorders, a 3rd of all days off work are caused by mental disorders, better access to therapy or psychotherapeutic drugs ->ppl can manage their mental health ->contribute to economy
correlation
an association between two co-variable use scatter gram to plot coefficient = strength positive if number =+X Negative if number = -X strength depends on closeness to 1 - if further = weak
Case studies
- idiographic
- detailed, in depth & longitudinal
- analysis of individual, group or institution.
- unusual events within a person
- qualitative data through case history collecting data from questionnaires, interviews etc..
case studies eval
+provides rich, detailed insight ->as more detail it can increase validity of data
+enables study of unusual behaviour -> some conditions are rare and this will help understanding of an individuals typical functioning
-prone to researcher bias ->uses subjective interpretation ->reduce the validity
-small samples -> studies only a small group/ individuals with unique charachteristics -> low external validity as we can’t generalise to whole populations
content analysis
ppl studied indirectly through communications
- type of observational research
- e.g. Spoken interaction.
content analysis -coding
coding - information is categorised into meaningful units
can involve counting up the number of times a particular word
thematic analysis
more qualitative
produce themes - more descriptive
can be categorised into broader categories
content analysis eval
+many ethical issues don’t apply -> material studied may already be in public domain ->no issues with obtaining consent.
+ flexible method ->produce both Quantitative & qualitative data -> can be adapted to suit the aims
-communication is studied out of context -> may attribute motivations to speaker that aren’t intended ->reduce the validity of conclusions drawn
-may lack objectivity -> threatens the validity of the findings
reliability
the measure of consistency, if a measurement is repeated and the sane result is obtained, then it is reliable
assesing reliability
- test then retest -> test same person twice -> results should be similar each time
- inter-observer, compare observations between two or more observers -> conduct a pilot study to test that observers are applying the same behavioural categories
- use a correlation -> should be +0.80 or more for reliability
improving reliability
- questionnaires -> rewrite open questions to closed, fixed choices -> less ambiguous.
- interviews -> improve training and try use same interviewer each time
- observations -> operationalise the behavioural categories =measurable, no overlapping - may be inconsistent judging.
- experiments -> use standardised procedures
validity
is the result legitimate?
whether an observed effect is genuine & represents what is seen in the real world.
internal validity
control within a study e.g. demand characteristics
external validity
generalising to other settings, populations or eras
ecological validity
whether findings can be generalised from one setting to another
-relates to the realism of the ppt’s task
temporal validity
findings should be consistent over time
e.g. Asch’s study lacks temporal validity because it was conducted during a conformist era in US history
face validity
- way to asses validity
- a test looks like it measures what it claims to do
- done by measuring the instrument or getting an expert to check
concurrent validity
whether findings are similar to those on a well-established test
improving validity
- experiements -> control group (the researcher is sure that changes in dv were due to effect of the IV) & standardised procedures
- Questionnaires -> use a lie scale to control for effects of social desirability bias -> use confidentiality to assure respondents of their privacy
- observations -> use good categories = well defined, operationalised
- qualitative research -> triangulation (multiple sources of evidence)
statistical tests
determine whether a difference/ or association found in an investigation is significant
Chi squared
- Test of difference, unrelated design and nominal data
2. Test of correlation, nominal data
Sign Test
- Test of difference
- related design
- nominal data
Mann-whitney test
- test of difference
- ordinal data
- unrelated design
Wilcoxon test
- test of difference
- ordinal data
- related design
Spearmans Rho test
- test of correlation
- ordinal data
Unrelated t test
- test of difference
- unrelated design
- interval data
Related t test
- test of difference
- related design
- interval
Pearsons R
Test of correlation
Interval data
Nominal data - level of measurement
each item can appear once in one category
no order
categories
ordinal data - lom
- numerical data, in ordered scale but intervals vary
- based on opinion, emotion and is abstract -> subjective
Interval data -lom
- numerical scales with equal units
- better than ordinal as more detail is preserved
Probability
the null hypothesis is accepted or rejected at a level of 0.05 or 5% probability of the event occuring
Significance
if not significant, the null hypothesis is accepted - ‘no difference/ correlation’
Type 1 error
null hypothesis is accepted & alternative hypothesis is rejected but null hypothesis is ‘true’
a false positive - optimistic as significant d/c is found when it =0
Type 2 Error
- null hypthesis is accepted, but the alternative hypothesis is true
- pessimistic error
- false negative
Errors & likilihood
- type 1 = if significance level is too lenient like 0.1
- type 2 = if significance is too strict like 0.01, potential values can be missed
sign test method
- enter pairs of related data into a table
- for each pair, score plus or minus
- = if item 2 is larger than item 1, + = if item 1 is larger than item 2 - S= the number of less frequent sign
- use critical value to calculate significance
scientific report sections
- abstract
- introduction
- method
- results
- discussion
- referencing
Abstract - sr
a short summary that includes aims, hypotheses, method, results and conclusions
Introduction-sr
- looks at relevant theories, concepts and studies
- follow logical progression, start broad then specific
Method -sr
- other researchers should be able to replicate study
- include the design +justifications, sample, sampling method, target population, apparatus and materials, procedure, ethics
Results - sr
-summary of key findings
-include descriptive statistics, inferential statistics
with the final outcome& any raw data.
Discussion - SR
- summary of findings in writing
- relationship of the results to previous research
- Limitations, suggestions to change
- wider real-world implications of the research
Referencing
- reference journal articles including author, date etc..
- reference books, authors, dates, publisher etc..
- reference web sources
features of a science
- paradigm & paradigm shifts
- theory construction
- falsifiability
- Replicability
- Objectivity
- Empirical method
Paradigms
Kuhn 1962 - a paradigm is a shared set of assumptions and methods, psych = pre-science -> lack of universally accepted paradigms
Paradigm shifts
- occur when there is scientific revolution
- researchers questions the accepted paradigm when there is too much contradictory evidence to ignore.
Theory construction
- a set of general laws/ principles that have the ability to explain particular events/behaviours
- need clear & precise statements to test
- deduction = deriving a new hypothesis from an existing theory
Falsifiability
- popper 1934, the key criteria of a scientific theory is its falsifiability
- should be used to test hypothesis and have the possibility of being proved false
- If cant be falsified & tested -> Pseudoscience
Replicability
- the findings of a theory must be shown to be repeatable across a number of different contexts
- replicability allows us to see the extent of which the findings can be generalised
Objectivity
- researchers keep a distance during research -> not allow personal opinions/biases to change the data
- methods with high control are the most objective
Empirical method
-data collection based on direct, sensory experience
-experimental & observational method = good examples
a theory can’t claim scientific unless it has been empirically tested
