Research Methods Flashcards
Experimental method AO1
extraneous - nuisance variables and dont vary with iv
confounding - change with iv so cant be sure if dv is due to iv or cv - must be controlled
Demand characteristics - cues from researcher giving aim away
Investigator effects - effect of investigators behaviour on dv
Research techniques -
Randomisation - use of chance to control for bias
Standardisation - use same procedures for all pps
Control groups - act as baseline and help establish causation
Single blind - pp doesnt know aim so demand characteristics lower
Double blind - pp and researcher dont know aim to reduce bias from demand characteristics and investigator effects
Experimental design AO1 and AO3
Independent groups - one group do condition a and one do b, randomly allocated - unrelated
+no order effects - pps only tested once, no practice/boredom controls confounding variables
+cant guess aim - pps only tested once - behaviour more natural
-pp variables - pps in two groups differ - may reduce validity
-more pps - more time spent recruiting - expensive
Repeated measures - same pps used in all conditions - related
+pp variables - person in both conditions same characteristics controls confound variables
+less pps - than independent - less time recruiting
-order effects - pps may do better or worse when doing similar task twice - lowers validity
-pps may guess aims - may change behaviour - lowers validity
Matched pairs - pairs of pps used but related to each other based on pp variables that affect dv, one member of each pair does one condition and other does the other- related
+pp variables - matched on variable - increases validity
+no order effects - pps only tested once - no practice/boredom - increases validity
-matching not perfect - time consuming, cant control all variables so may not address all pp variables
-more pps than repeated measures - more time spent recruiting - expensive
Types of experiment AO1 and AO3
Lab experiment - controlled, extraneous and confounding variables controlled
+evs and cvs controlled - effect of evs and cvs on dv lowered - actual cause and effect between iv and dv so high internal validity
+replicated cos standardised procedures - if same results then valid
-generalisability - artificial and pps aware theyre being studied - not natural and cant be generalised to everyday life - low external validity
-demand characteristics - cues that invite particular response - results may be explained by cues than iv
Field experiment - natural setting
+natural - pps more comfortable in own env - generalisable
+pps unaware of being studies - behave normally - increases generalisability and external validity
-hard to control cvs - observed changes can be due to cvs - hard to establish cause and effect
-ethics - no informed consent - invasion of privacy
Types of experiment AO1 and AO3
Natural experiment - iv varies naturally not by experimenter
+ethical - may be unethical to manipulate iv - only way for causal research
+high external validity - real life issues - findings relevant to real life experiences
-natural event rare - one offs - reduces opportunity for research limits scope of generalising
-pps not randomly allocated - iv preexists so researcher cant control whos placed in what conditon - cvs that arent controlled
Quasi experiment - iv based on pre existing difference, not manipulated
+high control - controlled conditions share strengths of lab - increases confidence about causal relationships
+comparisons - iv differs between ppl - comparisons between diff types of ppl can b made
-pps not randomly allocated - iv pre exists researcher cant control whos placed in which condition - pp variables act as a cv causing change in dv
-causal relationships not demonstrated - researcher cant manipulate iv - cant say if changes in dv due to iv
Sampling AO1 and AO3
Population - large group researcher interested in studying
Sample - smaller group taken
Generalisation - sample should be representative of pop so generalisable
Bias - certain groups may be over/under represented
Opportunity sample - most available ppl ie ask ppl nearby
+quick - convenient - just make use of ppl closest to u - popular meth
-biased - sample underrepresentative - drawn from very specific area - not generalisable]
Volunteer sample - pps select themselves ie by adverts
+pps wiling - know time and effort needed - probs engage more than ppl stopped on a street
-biased - pps may share traits - cant be generalised
Random sample - everyone has equal chance of being selected ie numbers placed into hat
+unbiased? - researcher has no influence over selection so free from researcher bias
-unrepresentative - random sampling can still produce biased sample - cant generalise
Systematic - pps selected using set pattern ie every nth person
+unbiased - first selected at random - objective
-time and effort - whole list of pop needed - might aswell use random
Stratified - pps selected according to their frequency in pop ie subgroups identified and percentages of subgroups are reflected in sample
+representative - characteristics of pop represented - generalisable
-not perfect - strata cant reflect all ways which ppl diff - complete representation not possible
Ethical issues AO1
Conflict between rights of pps and aims - bps code of conduct protects pps based on respect competence responsibility and integrity - ethics commitees weight up costs and benefits
Informed consent - permission from pps to take part, questions
Deception - misleading or withholding info so no informed consent - pps should be given debrief of true aims, details of witholded info, how their data will be used and their right to withhold data
protection from harm - pps shouldnt be at risk - right to withdraw, reassured their behaviour was normal, provide counselling if needed
Confidentiality - right to withhold info on ourselves - privacy respected - personal details protected, dont refer to pps using their names
Correlations AO1 and AO3
Association - strength and direction of association between 2 variables, correlations on scattergram w one co variable on x and one on y axis
Positive correlation - co variables rise/fall together
Negative - one rises one falls
Zero - no relationship
Correlation - influence of evs not controlled so maybe third untested variable acts as intervening variable showing cause
experiment - researcher manipulates iv records its effect on dv, in correlation, no manipulation of variables so cause and effect cant be established
+useful starting point - by assessing strength and direction, correlations provide precise measure of how two variables relate - may suggest future hypotheses
+economical - no need to manipulate variables and control - correlations less time consuming and expensive than experiments
-no cause and effect - correlations presented as causal but intervening variables might be cause
-flawed methods to measure variables - ie aggression score low in reliability as observation used - lowers validity
Observation AO1 and AO3
Seeing and listening to what ppl do without asking
+captures unexpected behaviour - ppl act differently to self report - more insight into normal behaviou
-observer bias - researchers interpretation affected by expectations - bias reduced with 2 observers instead of 1
Naturalistic - where target behaviour normally occurs
+high external validity - behaviour more likely to be spontaneous - more generalisable to everyday
-low control - of cvs and evs - difficult to see actual patterns
Controlled - some control of variables
+replicated - standardised - if results same then its valid
-low external validity - behaviour may be artificial cos of setting - findings not of everyday
Covert - pps unaware of being studied
+demand characterstics lower - behaviour natural - increases validity
-ethics - no informed consent, may not want to be watched - pps right to confidentiality affected
Overt - pps aware of being studied
+more ethical - gave consent - have right to withdraw
-demand characteristics - know being studied influences behaviour lowers validity
Participant - researcher becomes part of group being studied
+more insight - experiences same situation - increases validity
-loss of objectivity - researchers may identify with those studied - threatens objectivity and validity
Non participant - researcher separate from group being studied
+more objective - less bias cos researcher maintains objective distance - increases validity
-less insight - researched removed too far those being studied - lowers validity
Observational design AO1 and AO3
Behavioural categories - target behaviour observed and broken down into observable categorises clearly defined and measured - operationalised
-hard to make clear - categories shouldnt overlap hard to achieve ie smiling and grinning
-dustbin categories - dumped behaviours go unrecorded, all should be part of list
Time sampling - regular intervals
+lowers number of observations - than recording everything - more structured and systematic
-unrepresentative - may miss important details outside interval - might not reflect whole behaviour
Event sampling - behaviour recorded every time it occurs
+records infrequent behaviour - still pick up behaviours outside intervals which couldve been ignored in time sampling
-complex behaviour oversimplified important details unrecorded if too complex - affects validity
Self report techniques AO1 and AO3
Questionnaires - preset questions
+distibuted to lots - gather lots of data quick and researcher doesnt need to be present - cost effective - less effort involved
+respondents can be willing to open up - share more personal info than in interview - less social desirability bias than interviews
-responses may not be truthful - tend to present themselves in positive way - social desirability bias
-response bias - may favour a certain kind of response ie always agree so all respondents reply in similar way!!!!!1
Interviews - face to face interaction
Structured - list of pre determined questions in fixed order
+replicate - standardised - reduces differences between interviewers
-interviewees cant elaborate or deviate from topic - source of frustration
Unstructured - no set questions - but general topic - interaction is free flowing
+flexibility - points can be followed up - gain insight into interviewees views
-hard to replicate - not standardised - interviewer bias
Semi structured - list of questions but interviewers free to ask follow up when appropriate
Design of self report techniques AO1 and AO3
Closed questionnaires - respondent has limited choices
+easy to analyse - produce graphs and charts to compare - easy to draw conclusions
-respondents restricted - forced into answer that might not be true feelings - lowers validity
Open questionnaires - respondents provide own answers in words
+not restricted - answers provide detailed unpredictable info - more validity than statistics
-hard to analyse - wide variety of answers than quantitative - forced to reduce data to statistics!!!!
Interview schedule - list of questions interviewers need to cover - standardised - reduces interviewer bias
Quiet room - interviewee comfortble to open up
Rapport - begin with neutral questions to let pp feel relaxed
Ethics - remind that answers confidential
Trial run - small scale trial run before doing real - pilot study to find out if certain things wont work out so u can correct b4 acc spending time and money on real thing
Types of data AO1 and AO3
Quantitative - numerical
+easy to analyse - draw graphs - eyeball data and see patterns at glance
-oversimplified - ie rating scales to express feelings - individual meanings lost
Qualitative - words
+complexities - more detail and unpredictable info
-hard to analyse - lots of detail to summarise - hard to draw conclusions
Primary data - first hand data
+fits job - study designed to extract only data needed - info relevant to aims
-requires time and effort - plan and prepare - secondary data accessed within mins tho
Secondary data - collected by someone other than person conducting research e.g. websites
+inexpensive - already exists relevant info less effort less expensive
-poor quality - outdated or incomplete - challenges validity
Meta-analysis - secondary data combined from lots of studies
+more valid - much larger than individual samples - increases generalisability
-publication bias - researchers may leave out non significant results - biased as it only represents some data - incorrect conclusins drawn - less valid
Measures of central tendency AO1 and AO3
Mean - arithmetic average
+sensitive - includes all data - overall impression than median and mode
-unrepresentative - one very large or small number makes it distorted - median and mode not so
Median - middle value
+unaffected by extreme scores - median focused on middle value - more representative of whole data
-less sensitive - not all scores used and extreme values may b important
Mode - most common
+relevant to categorical data when data is discrete
-overly simple measure - many modes - cant describe data when theres more than one mode
Range
+easy to calculate - simple formula than standard deviation
-doesnt account for whole distribution - doesnt indicate if numbers closely grouped around mean or spread evenly - standard deviation better measure of dispersion in this respect
Standard deviation - average spread around mean - larger it is the more spread out data is
+more precise than range - includes all values - more accurate picture of overall distribution
-misleading - may hide some characteristics - extreme values not revealed unlike range
Graphs AO1
Tables - raw scores in columns and rows and summary paragraph below explaining results
Bar chart - discrete data categorised on x asis, height of each column represents its frequency on y axis
Histogram - bars touch each other - continuous data
Line graph - shows how something changes over time
Scattergram - correlational analysis - dot represents pair of related data
Normal distribution - symmetrical bell shaped curve w most ppl in middle
Skewed - lean to one side
Negative skew - leans to right most distribution towards right, mode is peak, median next on left, mean after
Positive skew - leans to left most distribution towards left, mode is peak, median on right and mean after
Statistical testing AO1
Significance - association between two sets of data not by chance - use statistical test - whether hypotheses should be rejected or accepted, whether diff/assoc due to chance or statistically significant
Probability - p value = 0.05 for psychology (5% or less that results due to chance/significant, 95% not due to chance/significant
SIGN TEST: Related design, Test for Difference, Nominal Data
Work out difference between two sets of data, ie 30-20 gives positive value, 20-30 gives negative value, add up total + and -, if difference is 0 ignore, S= no of less frequent sign, N= no of pps after deleting 0 values, one tailed = directional hypothesis - will be increase or decrease, two tailed = just looks for change, compare s value against critical values table
Null hypothesis states there is no relationship between two variables being studied (one doesnt affect other)
If S is less than/equal to critical value, reject null hypothesis, it is significant
there is correlation/difference and accept alternate hypothesis results are not due to chance
If ‘r’ in name of stat test used ie Spearmans ‘R’ho, value has to be greater/equal to critical to be significant
