W3 research methods Flashcards
(33 cards)
types of research
Qualitative
- non-numerical
- subjective experience (bottom-up)
- small sample
- collected in interview/group/natural
- flexible
for sociology
types of research
Quantitative
- numerical
- test hypotheses (top-down)
- large sample size
- collect in survey/experiments
- inflexible structure
error bars
represent the variability in scores
- standard error of the mean
- confidence interval 95%
measurement in quantitative research
some things easy to measure (height, income)
others are difficult (anxiety, intelligence)
- eg constructs of live not observable but inferred from observable behaviour
psycs methods and instruments to assess constructs indirectly
example - operationalisation of love
intensity of feelings - scale no love to overwhelming
satisfaction - eg 1 strongly dislike, 10 strongly love
frequency of expression
Types of variables
Scales of measurement
categorical + continuous
- nominal
- ordinal
- interval
- ratio
Nominal scale
scale categories preferences without order
fav flavours choc vanilla, mint
Ordinal scale
scale categories preferences with meaningful order
1. choc, 2 mint, 3 vanilla
Interval scale
scale has equal intervals
no true zero point
eg. ice cream enjoyment 0 strong dislike, 10 strongly like
0 doesnt mean no enjoyment
Ratio scale
scale has equal intervals and zero point
eg. # of scoops eaten
0 means 0
Measurement error =
2 types
minimise it means…
difference between observed and true score of variable being measured
1. random, 2. systematic
more reliability and validity
- random error
unpredictable factors that vary across measurement
eg. mood/weather means less hand hold
- systematic error
consistent biases in measurement
eg. some cultures less public display of love
assessing quality of a measure
Reliability =
consistent results and reproducibility of measurements across different instances
- similar result under similar conditions
not true score eg. im 75kg, it say 88,87,85
Validity
extent to which it accurately measures what is claims to measure
valid, accurate, reliable
eg. 75kg, it says 75,74,75.5
3 types of reliability
test-retest = consistent measurements over time
- internal = consistency across items in a test
- inter-rater = consistent across different observers
Surveys
example
public opinion by structured interview/questionnaire
- objective
- widely used
- gain representative sample of popn
- cost effective
eg. measure attitudes likert 1-5
Issues of survey research
- corrupted by social desirable responding
- differences in presentation influence responses
Experiments
manipulate predictor IV eg water
allow outcome DV to vary naturally eg growth
- but impossible to know everything that affects outcome, thus randomisation
True randomised experiments
cause and effect between variables
- randomly assign population to groups
- designs treatment (IV)
- requires control + experimental group
- analyse both outcomes
Between-subjects design
random popn
half to treatment B, and other half to B
- simple
- eliminates carry over
- shorter testing, less fatigue
Within-subjects
each participant is exposed to each treatment
- more statistical power
- smaller sample
- control of other variables like age
- efficient data collection
limitation of true randomised
monster study 1939
- non-consensual study on stuttering
- positive - encouraged
- negative - criticised speech
–> negative group developed speech problems
ethical + practical
Quasi experiment
lack randomisation, sacrifices control
- non-random assignment to groups
- researcher doesnt design the treatment
- pre-existing
- control groups not mandatory
