Research methods Flashcards
Ethics
Ethics are moral codes laid down by professional bodies to deal w problems during research
British Psychology Society (BPS) developed ethical guidelines for psychologists to follow when designing studies so p(s) r protected (principles for what’s acceptable/unacceptable):
-
Consent:
• u16s need parental consent to be apart of research
•p(s) must be told nature,aim, purpose of study n their role
•told they’ve right to w/d -
deception
•if deception used, informed consent can’t have been given if p didn’t get sufficient info about study
•researchers allowed to withhold info abt study if belief p wouldn’t behave naturally
•can deceive if deemed scientifically justifiable by ethics committee n if p wouldn’t object when deception revealed -
protection from harm
•BPS says psychs gotta protect p from physical/emotion harm n can’t expose em to more risk than they’d face in everyday situ -
Debrief/right to withdraw
•debrief should return em to state they were @ b4 study n explain what research involved
•must b told they can leave anytime during study n results can be withdrawn even retrospectively -
Confidentiality
•p shouldn’t be identifiable, data shouldn’t be linked to em n should be informed if info not fully anonymous
Dealing w ethical issues
Ethics committees often use cost&benefit analysis to determine if research if ethically acceptable.
Benefits may involve:
•ground-breaking nature of research
•research w irl app
Costs may involve:
•effect on p(s)
•reputation of psychology
-
Dealing w informed consent:
•p should b given consent letter w all info that may affect decision to participate
•p u16 need parental consent to partake
• Presumptive consent is to protect objective of study n involves asking a similar grp of ppl of study acceptable instead of acc p(s). If yes, then consent is assume
• General prior consent where p agrees to partake in number of studies including 1s w deception so consents to being deceived
• Retrospective consent asked for consent during debrief. must be allowed to withhold data if unhappy w data being shared -
Debrief:
•gotta be told true aim of study n details that weren’t given during it -
Counselling:
•given if p embarrassed or anxious after study -
Anonymity:
•needs to be maintained to protect personal details of p. Usually done w initials or no info at all eg case study
Variables
research involves changing n controlling variables j like in natural science
Variable= any “thing” that can be changed w/i an investigation. they’re generally used to determine if changes in one thing results to changes in another
-
IV:
variable that’s manipulated by researcher or varies over time -
DV:
•variable that’s measured/observed by researcher. Ideally in a study only IV should influence DV.
•b4 any results can b obtained, variables must be operationalised -
Operationalising:
clearly defining variables in measurable forms
•eg attachment, may investigate how parental interactions affect babies’ development.
gotta operationalise ‘development’ . could do this by assessing speech development eg how many words baby learns after certain time period?
may even operationalise ‘parental interactions’ which’s the IV. can measure this by amount of time spent w baby (in exams gotta operationalise variables for diff scenarios)
Factors affecting variables
Experiments can be contaminated by variables other than the IV, similar to control variables in natural sciences, these could be:
-
Extraneous variables (EV):
any variables other than IV that may affect DV if not controlled- external
eg SITUATIONAL VARIABLES:
room temp or age
or EXPERIMENTER VARIABLES:
unconscious bias/characteristics of experimenter -
Confounding variables (CV):
variable other than IV affecting DV but this varies systematically w the IV so can’t tell if change in DV cos of IV or CV
eg p(s) personality is systematically same every time so may affect behaviour every time -
Demand characteristics:
cues that reveal to the p(s) purpose of study so they do what experimenter ‘wants’ or the opposite -
Investigator effects:
any behaviour from investigator whether conscious/unconscious that affects DV. Could include things like design of study, selection n interactions w p(s).
eg
•The researchers expectations influences behaviour towards the participants, so they may respond with demand characteristics
•researchers expectations could result in p(s) asking leading qs
•p(s) may react to behaviour of an investigator and answer differently
Controlling research
single-blind study:
•deals w demand characteristics. •Researchers ensure p(s) have no way of knowing which condition they partaking in (often by deception) to ensure they behave naturally n don’t guess aim of study
Double-blind study:
•minimises demand characteristics/investigator effects so increases validity of responses obtained
• researcher n p(s) dk research hypothesis (eg assistant researcher carries it out)
-
Counterbalancing:
•minimises order effects
• half p(s) take part in condition A then B, n the 2nd half take part in B then A so balances order effects across the conditions -
Random allocation:
•avoids bias from p(s)/researcher in allocation of experimental/control grps
• allocation done randomly so individual differences can’t affect results much
eg if testing effect of sleep on memory n have all uni students in 1grp n all inactive ppl other grp then this’ll be CV so random allocation reduces effect of individual differences occurring -
standardisation:
•exact procedures given to all p(s) in study so they all experience same thing n therefore EV removed -
Randomisation:
•material presented to p(s) in random order to remove possibility of order effects
Pilot studies
small scale research to foresee any issues that maybe encountered in main study
allows researchers to identify potential problems w:
•procedures
•ethics
•materials
•design/method
So pilot study can rectify these if any n saves time n p
- pilot studies easier w some research but less viable w others eg natural/case studies as events r rare so wasteful to sacrifice a sample for this
Types of experiments
- lab
- field
- natural
- quasi
next slides will say meaning of each n pros n cons
Lab experiments
occur in controlled env where IV is directly manipulated by researcher n its effect on DV directly measured whilst maintaining control of EVs
FOR:
-
replication:
standardised procedures so easily replicable to check accuracy of findings -
Establishing cause n effect:
possible to est if IV caused change to DV -
High control:
CVs n EVs highly controlled so likely only IV affected DV therefore high internal validity
AGAINST:
-
Artificial:
may no measure irl behaviour so lacks eco validity -
Demand characteristics:
p(s) aware in lab study so may look to researcher/situ n guess aim which biases results. Therefore low validity -
Ethics:
deception often used so informed consent is difficult
Field experiments
occurs in natural env instead of lab n researcher manipulates IV n measures effect on DV
FOR:
- Eco validity higher than lab cos irl setting so is therefore more generalisable
AGAINST:
- low validity:
this is as researchers control of EVs n CVs reduce in field experiments
- Ethics:
if p(s) unaware being studied they can’t consent to it so mayb an invasion of privacy
Natural experiments
change in IV not done by researcher n would’ve occurred even if researcher wasn’t there.
called natural experiment cos IV changes naturally (not cos in natural setting- natural experiment can occur in lab too)
FOR:
- can conduct studies that wouldn’t have been possible due to ethical issues
- Higher generalisability cos involves study of irl issues and if conducted in lab then has same advantages as lab
AGAINST:
-
low reliability:
•can’t be replicated easily cos gotta wait for IV to naturally occur again so experiments are rare to conduct -
Can’t est cause&effect:
•bcz could be other variables affecting DV n can’t control CVs
Quasi experiments
IV not manipulated by researcher as they simply j ‘exist’ (is feature of participant eg age or gender)
•eg in Sheridan n King’s puppy shock experiment the IV was gender (54% men vs 100% women gave full shock)
FOR:
-
High eco validity:
less involvement from researcher cos of naturally occurring IV differences therefore can be applied to irl situs n has high external validity
AGAINST:
-
Participant allocation:
• can’t randomly allocate p(s) so can end up w unequal grps which threatens internal validity -
sample bias:
researcher has no control over p(s) that partake so samples maybe biased (eg on age/gender) so sample lacks pop representation
Observations
•means j watching closely.
•Many diff types each w advs n disadvantages.
• Observation studies suffer from observer bias- observers may interpret data according to their expectations, thghts our emotions which undermines objectivity
Types of observations:
- Controlled
- naturalistic
- participant
- non-participant
- covert
- overt
Controlled observation
behaviour observed in controlled conditions eg lab where some variables maybe controlled by researcher. p(s) aware being observed eg Ainsworth’s SSC
FOR:
-
High control
over EVs so high internal validity -
replicability
easy so results can be generalised
AGAINST
-
low external validity
results hard to generalise outside lab settings -
Hawthorne effect
p(s) adapt behaviour cos lnow being observed so leads to low validity
Naturalistic observation
Research carried out in natural setting instead of lab
FOR:
-
High eco validity
irl situs so results can be generalised
AGAINST:
-
Replicability difficult
cos carried out in natural env so reliability is low -
can’t est cause&effect
cos risk of EVs so conclusions hard to make
Participant observation
Researcher is a part of the group being studied
FOR:
-
high validity
richer insight into study cos researcher gets more detail about grps behaviour
AGAINST:
-
Lacks objectivity
•bcz researcher may start to associate w p(s)
•if covert, hard to take notes/record observations so has low reliability
Non-participant observation
researcher observes behaviour without being involved
FOR:
-
objective
not affected by subjective thoughts or emotions much
AGAINST:
- could lose rich data as researcher observes from distance
Covert observation
Participants are unaware of being observed
FOR:
-
High validity
bcz removes hawthorne effect
AGAINST:
-
Ethics q’able
since no informed consent
Overt observation
p(s) observed w consent
FOR:
-
Ethical
easy to take notes n p(s) aware being observed
AGAINST:
-
Hawthorne effect
p(s) may change behaviour cos know being observed which reduces validity
Self-report techniques
- Questionnaires
- Interviews (structured/unstructured)
Questionnaires
a set of written qs askedwhere p(s) can directly give info abt themselves
•useful for surveying thoughts, feelings, beliefs/behaviours
•can be done online, irl, posted or left in public places
Open questions:
have no set answer and p(s) can answer in own words (provides qualitative data)
Closed questions:
fixed choices of responses (provides quantitative data)
A commonly used scale is Likert scale which’s a 5 or 7 point scale asking p(s) how much they agree or disagree w suin (imagine it horizontal):
eg: How often do you brush your teeth?
• Always
•Sometimes
•Not sure
•Rarely
•Never
FOR:
- cost effective n ez to distribute
- can reach hella ppl
- data easy to organise n analyse/quantify
- researcher can’t influence answers if not present
AGAINST:
-
qs maybe misunderstood
bcz no one there to explain to p(s) -
Social desirability bias:
p(s) may alter behaviour to be seen positively -
Acquiescence bias:
p(s) tend to complete q’aire in similar way (ticking yes) too quickly without reading qs properly. This is a form of response bias n affects validity
Guidelines for good q’aire
-
Type of data
• do u want qualitative or quantitative data? this will determine if you ask close or open qs
•make sure qs are appropriate -
Ambiguity
•make sure qs are clear to understand -
Double barrelled qs
Avoid qs that contain 2 qs in 1 as p may wanna answer differently or may not understand -
Leading qs
•avoid leading qs that lead p(s) to the prefered answer
eg mothers are important in forming attachment w kids don’t you agree?
•and avoid qs open to interpretation
eg are u clever? -
Complexity
Use clear english and avoid using jargon that everyone may not understand
eg Do u suffer from retroactive interference in ur daily life?
Interviews
qs asked usually asked face2face where p can directly give info about themselves
-
structured interview:
pre-written set of qs that interviewer doesn’t deviate from n all p(s) answer qs in same order
Pros:
- ez to replicate
- objective n standardised
Cons:
- lacks rich details
-
unstructured interview
no set qs n works like a convo. there’s a general aim that certain topic will be discussed n interaction is free-flowing. P(s) encouraged to elaborate on their answers as prompted by interviewer
FOR:
- validity
-rich insights
AGAINST:
- subjective interpretation
- low inter-observer reliability
-
Semi-structured interview:
combo of structured n unstructured interview. has pre-set qs but respondent free to expand on answers as interviewer can ask follow up qs
Designing interviews
- schedule it
- should be standardised qs
- usually involved 1 participant n 1 interviewer tho may sometimes involve multiple interviews
- ask neutral qs to make p feel comfy
- use q checklist
- behaviour of interviewer may affect how respondent acts
Correlations
correlation:
•rs bw 2 or more co-variables
•they’re plotted on scattergram where 1 co-variable on x-axis n other on y-axis
•experiments manipulate the IV to measure effect on DV whereas correlation no manipulation of variables so can’t est cause n effect
-
Co-variables:
•variables investigated w/i a correlation eg height n weight
•indicates a degree of similarity or dissimilarity bw the 2 variables
•not referred to as IV or DV cos not establishing cause-effect rs but rather the association bw em -
positive correlation:
as 1 co-variable increases so does the other
eg taller u are more u weigh -
negative correlation:
as 1 co-variable increases the other reduces
eg more ppl in room means less space -
Zero correlation:
when no rs bw co-variables
eg no relation bw number of ppl in Ldn n amount of rain in Peru -
Correlation coefficient:
•The mathematical value showing the direction n strength bw the co-variables
• is bw +1 or -1
• +1= perfect positive correlation
• -1= perfect negative correlation
eg
+0.2= weak positive correlation
+0.8= strong positive correlation
-0.2= weak negative correlation
-0.8= strong negative correlation
Evaluate correlations
FOR:
-
Practical:
uses secondary data so p(s) not required n so time/cost efficient. -
correlational analysis is precise:
can tell researchers exact strength n direction of rs bw co-variables - can make predictions from it
AGAINST:
-
can’t est cause&effect
can say variables related but not that one caused other - results maybe hard to interpret
-
ethics
info can be hard to interpret so inaccurate conclusions maybe made eg so data can be misused by media perhaps
Case studies
•case studies involves a detailed study of a single individual, institution or event
•psychologists often turn to an individual case to study atypical behaviour n info can b used from a range of sources eg from the individual themselves or their family members
•many techniques can b used
eg interviewing individual or observing them
•case studies usually longitudinal as they follow the individual over a long period of time
eg Clive Wearing case study where they found his STM is damaged but LTM was strong showing that they’re separate stores which supports MSM
For:
- can gain detailed info n can give insight into certain rare behaviours eg living w memory impairment. wouldn’t be ethical to study such rare behaviours w experimental methods
against:
-
Cant generalise
case studies are unique to individual so can’t be idiographic n generalise the findings
eg can’t generalise findings from a patient w brain damage as all brain damage cases are different -
Ethics:
Need to maintain confidentiality of person so uses initials n sometimes hard to get informed consent for some patients
Aims n hypotheses
-
Aim:
•general statement of what researcher intends to study (highlights purpose of experiment)
•broader n less precise than hypothesis -
Hypothesis:
a clear testable statement stating rs bw variables being tested (predicts outcome) -
Null hypothesis:
•is what u assume is true during study (trying to disprove this)
•data collected will either back this claim or won’t
•If data don’t support claim, then go w alt hypothesis instead
•often null hypothesis is prediction there’s no rs bw variables tested -
Alternative hypothesis:
•if null hypothesis rejected, this is the accepted alternative (what u acc wanna prove)
•eg if null hypothesis was variables aren’t linked, then alt= linked -
Directional (1 tailed) hypothesis:
•predicts outcome/direction of rs bw the variables
•often used when previous research findings suggest way research will go
eg exercise lowers risk of health problems -
Non-directional (2 tailed) hypothesis:
•doesn’t state direction of outcome for variables tested
•often used when no previous research or if previous research findings contradictory
Sampling
Population:
is the grp of ppl who are the focus of researcher’s interest from which a small sample is drawn
Target pop:
grp of ppl who share characteristics that researcher wants to draw a conclusion from. For cost n practicality reasons, as target pop is too large sample drawn from it
Sample:
grp of p(s) that take part in study. This is drawn from target pop n presumed to b representative or target pop
Sampling technique
is method used to select samples.
These mainly are:
- random sample
- systematic sample
- stratified sample
- opportunity sample
- volunteer sample
Random Sampling
where members or target pop have equal chance of being selected- usually by lottery method eg names from hat or computer randomiser
For:
- Unbiased as researcher has no control of who’s selected therefore CVs n EVs equally divided bw grps meaning high internal validity
against:
- time consuming as complete list of target pop maybe hard to obtain
- Unrepresentative sample possible as no guarantee of picking p(s) that have traits you wanna investigate
Systematic sampling
where every Nth member of target pop selected. A sampling frame(list of ppl in target pop) n then a sampling system’s nominated (eg every 7th person)
For:
- Unbiased bcz researcher has no control over who is selected
Against:
- time consuming to do n maybe not fully free from bias unless number selected for sampling is random
Stratified sample
represents a composition of people in certain subgroups(strata) within the target population
• researcher identified diff strata that makes up population
•then proportions needed for sample to be representative is worked out
eg in Manny:
50% support utd
40% support city
10% other
so if sample number is 10 then 5 gotta support utd , 4 citu n 1 supporting other. each of these would b randomly selected from larger grp of fans of their team
For:
- Unbiased as members randomly selected n sample represents strata composition better
AGAINST:
- individual differences aren’t fully accounted for may not be 100% representative of stratified pop
Opportunity sample
researcher selects anyone who’s readily available n willing to partake. usually asks whoever is around n convenient to ask eg if researcher works at uni may ask students to partake
For:
- Convenient bcz saves time n p bcz list of members not required
Against:
- Unrepresentative bcz sample drawn from v specific area so can’t generalise results to target pop
- Researcher Bias possible as researcher chooses who participates
Volunteer sample
where participants pick themselves to be apart of research. Researcher may place ad eg on noticeboard that volunteers are needed n p(s) may simply just respond yes or no
for:
- less time-consuming cos easy to do n less effort for researcher
against:
-
Volunteer bias
as asking for volunteers may attract certain ‘profile’ of ppl or someone affected by matter being studied which would make results less valid n not generalisable
Experimental design
experimental design is diff ways in which p(s) can b organised in relation to the experimental conditions
There’s 3 types:
- independent grps
- repeated measures
- matches pairs
Independent grps design
diff p(s) in each condition n the difference bw these grps would then be compared.
For:
- No order effects
- p(s) less likely to guess aim
against:
- Individual differences may become CV cos no control on participant variables
- Costly as twice as many p(s) needed to have equal number in each grp to produce the data
Repeated measures design
Same p(s) take part in both conditions then the results from each grp compared to c if there’s difference
for:
- Individual differences removed n therefore can be compared ‘like-to-like’
- cost effective cos same p(s) used in all conditions
against:
- Order effect v likely as doing many conditions can lead to boredom/fatigue (can be minimised w counter balancing)
- Demand characteristics more likely cos if they take part in all conditions likely to guess aim
Matched pairs design
pairs of p(s) r first matched on some variables that may affect DV. then the pair is split up into diff conditions
For:
- No order effects as p(s) only take part in one condition so no need for counterbalancing
Against:
- cost n time consuming to match pairs to each condition n esp if pilot study needed
Observational design
A key influence on observational design is how researcher wishes to record the data:
-
Structured observation:
scientific method, where behaviour simplified into categories
Structured observation follows a systematic sampling method to record behaviour:
• Event sampling involves counting how many times an event occurs(behaviour)
• Time sampling involves recording behaviour w/i pre-established time frame
eg record what target participant is doing every 15secs
•Structured obo produces quantitative data so analysing n comparing data straightforward -
Unstructured observation:
note everything that’s observed (gives rich detail)
•unstructured obo produced qualitative rich data but can suffer observer bias where behaviour is recorded subjectively or not at some instances so low in reliability -
Behavioural categories
•is operationalising as target behaviour broken down into components that r observable measurable.
•behavioural categories should be precise, measurable n exclusive n shouldn’t be open to interpretation eg falls asleep, or eats
Peer review
assessment of scientific work by other specialists in same field to ensure research intended for publication is of high quality
peers should be unknown to the author of that research n conduct their review objectively
Aims of peer review:
• decide if should allocate funding for research or not
• validate quality n relevance of research n ensure all research method principles r adhered to
• suggest amendments or improvements before publishing or to withdraw if necessary
Criticisms of peer review:
-
Anonymity
of reviewers maybe used to criticise rival researchers to limit their research funding -
publication bias
as publishers tend to produce ‘headline-grabbing’ findings n portrays their journal in positive light n anything that’s not headline grabbing maybe dismissed -
Burying groundbreaking research.
•peer reviewing process may suppress theories to maintain their mainstream theory n favour work that maintains status quo
• can slow down publication of valuable research which could be put to good use
Implications of psychological research for economy
•one of wider concerns of psychology other than increasing our general understanding of its implications for the economy (how does it value eco?)
Eg:
- Attachment:
Bowlby first stated kids can only form secure bond w mum n then recent research suggested fathers can also be as affective as mums in caring for kid. This means both parents can support child n work which in turn means pay more taxes. Now norm mums earn high income n both parents work n contribute to economy - Psychopathology
Absence from work costs eco hella p. patients w mental disorders can be treated w drug therapies like SSRIs or counselling meaning they can manage condition effectively n return to work so psychologically research into problems like depression or ocd has economic benefits
Reliability
reliability refers to extent to which a test provides consistent findings
•problems w reliability can occur when researcher tries code complex behaviour using a small number of categories eg record acts of aggression by a person
•having one observer for that could produce subjective judgement so more observers provide ratings for behaviour. ratings then compared to give a measure of inter-observer reliability
-
Inter-observer reliability:
measures internal reliability so may have multiple observers recording data n their records are correlated. if they agree then results reliable. if suin has high correlation, eg 0.8, there’s a lot of common ground bw observations so we’d say it has inter-observer reliability -
Test-retest method
another way to establish reliability. Here, we give same test to p(s) on separate occasions
eg give IQ test but this maybe affected by fact u have no sleep so give again another time n if it has same result then can say high reliability
Internal reliability is talking about internal consistency ie was procedure carried out same way each time
eg IQ test contains diff sections of equal difficulty, so p(s) should score consistently across each section
External reliability refers to how consistent results are over time regardless of when it’s used eg see test-retest method
• improving reliability can be done by increasing objectivity n making less subjective as possible n leaving less down to individual judgment
eg by operationalising which ensures we’re using same variables n doing experiment same way
or using standardised procedures which would make it internally n externally reliable so everything done is properly detailed n stays same each time it’s done
Validity
refers to extent study measures what it set out to measure
eg IQ test w only maths qs can’t be said to measure general intelligence so low validity
-
internal validity:
is having tight controls over EVs leads to higher internal validity bcz likely measuring what u say u are eg lab experiments -
external/ecological validity:
extent to which results can reflect to irl behaviour -
Face validity:
looking at test n seeing extent to which it looks like it’s testing what it claims to test on face of it -
Concurrent validity:
extent to which test produces same results as another est’d measure eg 2 IQ tests should produce same results but if has diff results then low concurrent validity -
Predictive validity:
testing the validity by checking if it can predict future behaviour
eg I did well on my GCSEs so i will do well in my alevels n if this is true then high predictive reliability -
Temporal validity:
extent to which the results of a test can be generalised across different times eg is it true for now or for past n future times? -
Populational validity
whether results can be generalised to larger pop or j the sample
eg maybe culture specific or age specific n not apply to everyone as a result
Improving validity
•internal validity can be improved w tighter controls of EVs n ensuring high reliability in procedures n measurements taken
•external/eco validity can be improved by developing realistic tests n using natural settings eg conducting research in real world
Features of science
-
Objectivity
crucial in science. Researchers should remain totally free of opinions n bias in their investigations ie not influenced by their personal feelings. This is so all sources of bias are minimised n subjective ideas eliminated. Pursuit of science implies facts will speak for themselves even if not want researcher hoped for -
Empirical methods
Refers to data being collected thru direct observation. Direct testing only way to confirm things. must have identifiable variables n control all EVs that can effect ability to establish cause n effect. Empirical evidence based on data not opinion n so experiments done carefully n reported in detail so it can be repeated n verified -
Replicability
linked to reliability of findings. refers to if methods n findings can be repeated by same ppl or others to c if results are similar. findings that can’t be replicated not generalisable so can’t make conclusions. if results same repeatedly then reliable n can be used to est a scientific theory. That’s why reports must be detailed so it can be replaced to check for validity n reliability -
Theory construction
based on idea that findings of empirical research, theories can be made to help us predict behaviour. this is linked to predictability n should be aiming to predict future behaviour from research findings -
Hypothesis testing
want to test our hypothesis (statement made at beginning which serves as a prediction for study-derived from a theory). There’s diff types of hypotheses (null n alternative) both of which need to be operationalised n unambiguous so they can be tested
-
Falsification
• Popper said theories are abstract n it’s only possible to disprove them
eg can’t prove every swan in world is white cos can’t c every in world n only takes 1 black swan to disprove it
•therefore, we begin with a null hypothesis which is worded to disprove the theory, which’s then accepted or rejected. If we falsify the theory, we accept the alt hypothesis.
•a benefit of falsification is avoids confirmation bias where researcher looks for what they want you prove n ignore anything contrary. this can reduced by starting off w null hypothesis -
Paradigms
• Kuhn claimed that science consists of an accepted paradigm. Paradigm is a set of principles n methods which define a scientific discipline. It’s a set of norms that tells a scientific how to think n behave n all scientists accept this. • Paradigm shift is where there is a change in beliefs n behaviour as new theories challenge the dominant paradigm n questions the basis of it till the new theory is accepted
eg scientists used to believe earth is centre of universe to accepting the sun is n all scientists work on this basis now
• psychology doesn’t have any major paradigms so arguable if it is a science, instead there are diff approaches eg behaviour, cog, bio which co-exist
Reporting psychological investigations
-
Title page
•indicates what studies about n must include variables under testing
eg the effects of hunger(IV) on reaction time (DV) -
Abstract:
concise summary of the report n should briefly tell reader the following without having to read whole report:
•aim n hypothesis
•method (design, p(s), procedure eg how data was recorded, major findings
•implications of findings -
Introduction:
•explain purpose of study n where hypothesis came from (provide rationale).
• should have funnel structure (start broad n become specific- psychological literature should logically follow into the aims n hypothesis)
eg highlight broader topic, then explain theoretical framework, then summarise previous studies, then rationale (why ur doing it eg overcomes a flaw in previous research), then hypothesis -
Method
report how research was carried out: (assume reader has no knowledge of what u did n ensure they can replicate)
•participants (how many, how sample obtained, etc)
•design (experimental design n potential problems w it, IV n DV, how EVs controlled eg counterbalances)
•resources (list materials used eg questionnaire, word lists etc)
•procedure (describe precise procedure followed in detail so can be repeated eg what happened each time p took part, how data recorded etc) -
Results
can be reported in descriptive statistics (tables, graphs, charts) or inferential statistics(doing statistical tests on data). Avoid interpreting results n don’t use raw data eg report numbers 2dp, whole number for %, when reporting 95% Confidence intervals, give upper n lower limits im square bracket
•results section needs to state Confidence intervals for the IV (95%), name of test chosen say why test chosen eg bcz testing correlation
•include results of study (eg stats,
magnitude n direction of results) -
Discussion
•explain findings n implications n relate to hypothesis
•limitations of study n how can be improved (if reliable, leave CVs out as doubts ur own research)
•rs to background research (relate results to previous studies n of ur findings support theirs) -
References
list of all sources used during study when referring to name n date of psychologists eg books, articles,websites etc
•allows reader to c where info from n where theories in report from
•report in alphabetical order
Types of data
- qualitative
- quantitative
- primary
- secondary
Qualitative data
•Data expressed in words n is non-numerical
•may take form of written description of thoughts/feelings/opinions or written account of what researcher observed.
•qualitative methods of data collection r those concerned w interpretation of lang (from interview or unstructured obo) n collected to express the quality of suin
eg if wanna know why ppl commit crime, we may use qualitative, confidential, unstructured interview which will produce qualitative descriptive data
For:
- provides rich detailed understanding of how mind works.
The data tells us something more about how people feel about the world in which they live, and why they behave as they do - so has high external validity bcz helps us understand meaning behind behaviour
Against:
- hard to organise n analyse bcz harder for researcher to categorise ppls responses (can’t easily make graph from responses for eg)
- relies on subjective interpretation which may be subjective to bias so not generalisable
- time consuming/costly- than quantitative since researchers looking at info in much more depth. Neither type of data is superior n both have uses so both maybe used
Quantitative data
•data that’s expressed numerically. The data’s usually collected as individual scores eg as an amount of words p was able to rmbr or rate or percentage
•the data can easily be converted into a graph or chart etc
For:
- ez to measure n analyse so comparisons can be made easily
- objective n not open to bias as collected simply by counting suin
Against:
- narrow tells us v lil about suin eg why ppl behave a certain way
lvls of measurement
Quantitative data can be:
-
Nominal data:
•a frequency count for data in separate categories
•v simple n don’t tell us no more than headcount
eg in class of 10, 5 support utd, 4 support Chelsea, 1 supports West Ham -
Ordinal data:
•Data that can be ordered and/ ranked
•Tells us nuin about interval bw each order (eg how much more did they prefer pasta over salad?)
eg asking someone to list their fav foods in order of preference may say pasta then salad then chicken -
Interval data:
•measurements taken on a scale where each unit is same size
•most precise type of data
•eg on a thermometer, there is a 1° differnce bw each line
or 12hr clock there’s 1hr differnce bw each each number like 9 n 10
Primary data
info personally obtained by researcher
eg by observation interview or questionnaire
for:
- authentic n fits purpose of research
against:
- time consuming n costly- requires planning n resources
Secondary data
Collecting info that already exists (done by other researchers or gov reports)
For:
- Timely n cost effective as minimal effort required to collect data
Cons:
- variation n accuracy info maybe outdated or incomplete n lack validity
Meta-analysis
•process of combining results from from other studies on particular topic to give an overview.
•studies used involves same research qs n use comparable methods
•this may involve qualitative review of conclusions or a quantitative analysis of results which produces a effect size(where the DV of the meta-analysis gives an overall statistical measure of rs bw variables across a number of studies)
For:
- creates larger more varied sample which means results can be generalised across larger pop
Cons:
-
Prone to publication bias:
researcher may not select all relevant studies n may leave out studies. This means conclusions of meta-analysis will be biased bcz it’ll lead to unrepresentative meta-analysis conclusion
Descriptive statistics
Descriptive statistics is the use of summary statistic (eg graphs n tables) to identify trends n analyse sets of data. It includes measures of central tendencies.
Central tendency is general term for any measure of average value in a set of data. 3 measures:
-
mean:
•calculated by adding up all values in set of data n then dividing by number of values
•most sensitive measure of central tendency as it uses all of data.
•can be distorted by extreme values -
median:
•middle value of scores when arranged in ascending order
•most suitable for ordinal data
•not distorted by extreme values as doesn’t use all data (not as sensitive as mean but can be distorted by small samples) -
mode:
•most frequently occurring value in set of data
•not affected by extreme values
•not useful when there’s many or no modes
Measures of dispersion
measures of dispersion is the measure of spread/dispersion of scores in set of data (how far they vary from each other. 2 ways to calculate this:
-
Range:
•difference bw highest n lowest value in set of data
• +1 is added to mitigate the margin of error
•only accounts for 2 most extreme values in set n maybe unrepresentative of data n doesn’t indicate if numbers closely grpd or spread out -
Standard deviation
• a single value that tells us how far scores deviate from the mean
• the larger the SD the greater the dispersion w/i set of data
•small SD tells us data closely grpd together
•precise as it includes all values before final calculation
•can be distorted by extreme skies or may not even indicate presence of extreme values
Presentation n display of quantitative data
-
Tables:
data represented thru raw scores or descriptive statistics -
Bar charts
•used when data divided into categories- discrete data
•frequency of each category on y-axis -
Scattergrams:
•displays the correlation bw 2 co-variables
•either co-variable occupies x or y-axis -
Histograms:
•graph displaying frequency but the area of bar represents the frequency (
•x-axis must start at true zero n scale is continuous -
line graphs:
•data represented by points connected showing how given variables change over time- continuous data
•may include straight line of best fit or curve of best fit
Distributions
-
Normal distribution:
•a symmetrical spread of data that forms a bell-shaped pattern. mean, median, mode all located in middle at highest peak
•most ppl located in middle (average) n few ppl on extreme end (left is low performer n right is high performer)
•tails of the curve extend outwards n never touch the x-axis as more extreme scores are always theoretically possible -
Skewed distribution:
a spread of frequency data that isn’t symmetrical as data clusters to one end
• positive skew:
most of contribution concentrated on left side n tail is extending to right side
(eg hard test n hella ppl performed bad n few performed good)
• negative skew
data clustered to right side n tail extends to left side
(eg easy test hella ppl did good n few did bad)
•in skewed distribution curve, mode remains highest point of peak n then median. mean is dragged to either left or right and depending on where skew curve goes (bcz mean affected by extreme values so indicates extreme movement). Mode n median don’t include all values when calculated.
Data Analysis
- thematic analysis (qualitative data analysis)
- content analysis (quantifies qualitative data)
Thematic analysis
Qualitative data is difficult to analyse objectively so is sometimes seen as having limited use.
Thematic analysis:
•form of qualitative analysis
•involves making summaries of data n identifying key themes/categories
For:
- preserves detail in data
- data can be summarised
against:
- decisions can be subjective n researchers maybe biased
- time consuming to summarise n categorise
This means thematic analysis is not objective
Content Analysis
Content analysis often used in observations n can b used to turn qualitative data quantitative.
• in a content analysis the research looks at the data to create a suitable coding system
• after analysing the data, researcher tallies any common themes to c how often it occurred
•this summarises qualitative data
into quantitative which makes the data easier to analyse
for:
-
high eco validity
bcz based on obos of what ppl acc do eg books ppl read
when these sources are accessible other content analysis can be repeated which can check the reliability of the observations
against:
-
risk of observer bias
diff observers may interpret results differently so observations maybe subjective. This means content analysis is not objective or valid
Inferential testing
-
Inferential testing:
•use of statistical tests to insert differences bw grps/samples n test if difference is significant or not
•helps draw a conclusion about pop from samples drawn
• sig difference indicates difference bw grps acc due to IV on DV n UNLIKELY by chance
•thus a sig or not sig difference indicates if null hypothesis should be accepted or not -
Statistical testing:
•provides a way of determining if hypotheses should be accepted or no. It indicates if rs/difference bw variables are sig (meaningful) or occurred by chance -
Probability:
•refers to likelihood of an event occurring
•In psychology, chosen lvl of significance usually 5% or p= 0.05 (determines if difference or rs bw variables is acc sig)
•some psychs may wanna replicate study w more certainty so use p= 0.01 or 1% esp if it involves human cost eg new drug therapy tested
there can only be 3 possible lvls of significance:
• p=0.05 means 5% chance difference/rs in study was by chance (researcher 95% confident results cos of manipulation of IV n not chance)
• p _<0.05 means 5% or less than 5% chance difference/rs in study by chance (researcher at least 95% confident results cos of manipulation of IV n not chance)
• p<0.05 means less than 5% chance difference/rs in study occurs chance (researcher more than 95% confident results cos of manipulation of IV n not chance)
Sign test
statistical test to analyse differences in scores bw related items eg same p tested twice
when to use sign test:
•it’s a test of difference so not suitable to find correlation bw diff variables
•data must be Nominal (categorised)
•data can b from Repeated measures design(each p tested twice)
info needed to carry out sign test:
• calculated value (worked out for you)
• critical value (given in critical value table)
• N which’s number of p(s) in study
• if hypothesis 1 tailed or 2 tailed to determine lvl of significance
Those last 3 bullet points allow you to locate critical value for ur data. For the sign test, calculated value gotta b equal to or less than critical value for results to be significant
sign test: worked example
-
What’s the hypothesis?
•is it 1 or 2 tailed? -
Is data nominal?
• yes, then use sign test -
Work out the calculated value of S
•(holiday eg) difference bw scores before n after of each p happiness calculated. If 2nd value higher then put + n if lower then - in the sign column.
• symbol for calculated value for sign test is S n is calculated by adding all plus signs n all minus signs. S is the smaller value of both eg in holiday example theres 3 minus signs n more plus signs so calculated value is 3 -
Find critical value of S using critical value table (provided)
• will require N(number of participants).
• leave any value where difference is 0 eg in holiday example there’s 14 p(s) but 1 had no difference so final N= 13
• Table has sections for if study is 1 or 2 tailed n pick whichever it is n look at values at 0.05 lvl of sig bcz that’s the lowest we prepared to accept -
Is it significant?
• for sign test value to be significant, calculated value of S gotta be EQUAL TO or LESS THAN critical value
how to state conclusion:
“at 0.05 lvl of sig, since the calculated value of 3 is equal to critical value of 3, we must accept the alt hypothesis that ppl are happier after holiday than before a holiday n reject the null hypothesis”
Probability
we use 0.05 lvl of significance or p_<0.05 meaning there’s 5% or less chance that results were due to chance n not cos of change in IV. basically means we 95% sure results due to IV
•if research sensitive, may use more stringent lvl of significance eg 0.01 meaning sure results at least 99% due to IV
Type 1 and Type 2 errors
-
Type I error
•when you falsely REJECT null hypothesis n alt accepted when shouldn’t have been
eg covid test shows ur pos but u don’t have covid.
•we assume results due to IV not chance
•significance lvl gives the probability of this happening which’s why sig lvls small -
Type II:
•null hypothesis falsely ACCEPTED n alt hypothesis rejected when shoulda been accepted
eg covid test shows ur neg but u acc have covid
•assume results due to chance but they were due to IV
•if the probability lvl set to lenient eg 10% then likelier to make type1 error n wrongly accept alt hypothesis
• probability lvl too stringent eg 1% likelier to make type2 error n wrongly accept null hypothesis