Research Methods Flashcards
aim
what the researcher intends to investigate
hypothesis
statement stating relationship between IV & DV
state the 3 types of hypothesis
null, directional, non-directional
null hypothesis
nothing will happen
directional hypothesis
1 tailed, one specific group will do better than the other
non-directional hypothesis
2 tailed, predicts something will happen but not ‘direction’ of the effect
independent variable
manipulated by researcher so DV can be measured
dependant variable
measured by researcher
what do you have to do once you have made the aims and hypothesis
operationalise variable e.g. turn UV & DV into something we can measure
name all research issues
extraneous variable, confounding variable, demand characteristics, investigator effects
research issues - extraneous variable
variable affecting DV
additional/unwanted - should be identified + have steps taken to minimise effects
doesn’t vary systematically with IV
research issues - confounding variable
varies systematically with IV so can’t tell is change in DV is due to IV or varaible
research issues - investigator effects
effect of investigator’s behaviour on research outcome
e.g. leading questions
Coolican: include expectancy effects + unconscious bias - actions of researcher related to study design (selection of PPs, instructions)
research issues - demand characteristics
cues from researchers / situation that may be interpreted as revealing purpose
act in way they think expected + over-perform (please-U effect)
underperform to deliberately sabotage (screw-U effect)
research issues (combat) - randomisation
control investigator effects through use of chance methods to reduce researcher’s unconscious biases
research issues (combat) - standardisation
using exactly the same formalised procedures / instructions for all PPs in a research study
case studies
in-depth investigation, description, analysis of single individuals, group, institution, event
tend to take place over long time (longitudinal)
content analysis
research technique enabling indirect study of behaviour by examining communication that people produce e.g. in texts, emails, TV, film
aim to summarise & describe communication in systematic way so overall can be drawn
coding
stage of content analysis - categorise large sets of info into meaningful units
may involve counting up number of times particular word/phrase appears to produce quantitative data
thematic analysis
inductive & qualitative approach to analysis that involves identifying implicit/explicit ideas within data
themes often emerge once data been coding
evaluate case studies
+ rich, detailed insights may shed light on unusual + atypical forms of behaviour - preferred to more ‘superficial’ forms of data from experiment/questionnaire
+ contribute to understanding of ‘normal’ functioning e.g. HM significant as it demonstrated ‘normal’ memory processing
+ generate hypotheses for future study + 1 solitary contradictory instance may lead to revision of entire theory
- generalisation when dealing with small sample sizes
- information in final report based on subjective selection & interpretation
- personal accounts prone to inaccuracy & memory decay (low validity)
evaluate content analysis
+ useful - gets around ethnical issues
+ most material already exists within public domain - no issues obtaining permission
+ high external validity
+ flexible - produce both qualitative & quantitative data
- people tend to be studied indirectly so communication they produce usually analysed outside context within which it occurred
- researcher may attribute opinions
- lack objectivity especially when more descriptive forms of thematic analysis employed
assessing reliability
- test retest: questionnaire, testing over time, correlation -0.8+
- inter-observer reliability: multiple observers - overcome bias 0.8+ pilot study
assessing reliability - test retest
method assessing reliability of questionnaire/psychological test by assessing person on 2 separate occasions
shows to what extent test produces same answers i.e. is reliable
must be sufficient time between test & retest to ensure PP can’t recall answers but not so long attitudes, opinions, abilities changed
assessing reliability - inter-observer reliability
extent to which there is agreement between 2+ observers
measured by correlating observations of 2+ observers
(total number of agreements)/(total number of observations) >+0.80 data have higher inter-observer reliability
issue: everyone has own unique way of seeing world - relevant observational research as researcher’s interpretations may differ - introducing subjectivity, bias, unreliability
reliability
how consistent findings are - measuring device said to be reliable if it produces consistent results
improving reliability - questionnaires
test-retest method
if produces low test-retest reliability may require some items to be ‘deselected’/rewritten (if questions complex may be interpreted differently) solution replace some of open questions with fixed choice alternatives
improving reliability - interviews
ensure reliability best by use same interviewer - if not possible interviewers must be properly training so no asking leading/ambiguous questions
easily avoided in structured interviews where interviewer’s behaviour more controlled by fixed questions
improving reliability - experiments
lab experiment reliable by struct control on procedure
precise replication of method rather than demonstrating reliability of findings
improving reliability - observations
reliability improved by making sure behavioural categories properly operationalised + measurable + self-evident - categories shouldn’t overlap + all possess behaviours should be covered
if categories not operationalised well different observers have to make own judgements of what to record + end up with differing and inconsistent records
validity
extent observed event genuine (does it measure what supposed to) (can it be generalised)
face validity
a measure scrutinised to determine whether it appears to measure what its supposed to
concurrent validity
extent psychological measure relates to existing similar measure
ecological validity
extend findings can be generalised to settings/situation
form of external validity
if task used to measure DV is not ‘like everyday life’ (i.e. low mundane realism can lower eco valid)
temporal validity
extend findings can be generalised to other historical times
form of external validity
internal validity
refers to whether effects observed due to manipulation of IV and not another factor
major threat to IV is if PPs respond to demand characteristics e.g. Milgram ‘played along’ + didn’t believe they were administering shocks
external validity
relates to factors outside investigation e.g. generalising to other settings, populations, eras
ecological
validity - qualitative methods
high eco valid than quanti (less interpretive) because depth & detail associated with case studies and interviews (better reflect PPs reality)
validity further enhanced through triangulation - use number of different sources as evidence e.g. interviews, diary, observation
type 1 error
reject null hypothesis when we shouldn’t - if probability level too loose/lenient (e.g. 0.1 an ‘optimistic error’
incorrect rejection of true null hypothesis (false +tive)
type 2 error
accepting null hypothesis when we shouldn’t - if probability level too tight/stringent (e.g. 0.01 a ‘pessimistic error’
failure to reject false null hypothesis (false -tive)
levels of measurement
quantitative data classified into types/levels of measurement
e.g. nominal, ordinal, interval
chi-squared
test for association (difference or correlation) between 2 variables/conditions
data should be nominal level using unrelated design
Mann-Whitney
test for significant difference between 2 sets of scores
data should be ordinal level using unrelated groups
Pearson’s r
parametric test for correlation when data at interval level
Related t-test
parametric test for difference between 2 sets of scores
data must be interval with related design
sign test
statistical test analyse difference in scores between related items i.e. same PPs tested twice
nominal
spearsman’s rho
test for correlation when data at least ordinal level
unrelated t-test
parametric test for difference between 2 sets of scores
data interval with unrelated design
Wilcoxon
test for significant difference between 2 sets of scores
data ordinal level with related design
experimental design
related design: repeated, matched
unrelated design: independent