Chapter 1 Research Methods Flashcards
(19 cards)
Independent Vs Dependent Variables
Independent: hypothesized effect
Dependent: hypothesized cause
both in experimental research. both causal
Predictor vs Outcome variables
predictor: hypothesized basis for prediction
outcome: what you’re trying to predict
both in correlational research, predictive variables.
Independent/Dependent vs predictor/outcome
independent and dependent are outcome and predictors bc they suggest causality
predictor and outcome NOT independent/dependent bc they suggest correlation and not causation
categorical variables
definition: distinct categ
* binary: 2 categories, also nominal (M/F)
* nominal: 2+ categories (religion)
* ordinal: categories with logical order (grades, race place)
category: group whose membership is based on same criteria
continuous variables
def: entities get a distinct score and is infinitely divisible
* interval: equal intervals represent equal difs (hours)
* ratio: same but scale makes sense (absolute zero)
the research process
- data
- initial observation/question
- theory
- generate hypothesis (idenitify variables)
- collect data (measure variables)
- analyze data (graph, fit model)
- back to theory
measurement error
discrepancy bw # we use to represent something & the actual value.
validity
whether it measures what it intends to measure
- content: do questions relate to construct
- ecological: reflects the real world
- external: shows results similar to what other related constructs would
- concurrent: other measure taken at the same time
- predictive: taken in future
reliability
produces same results under same conditions
- test-retest: see if correlated
- inter-item: correlation bw items
- inter-rater: if 2 raters agree
cronbachs alpha: measures inter-item
reliability and validity importance
- typically need reliability to be valid
- but, constructs can be inherently unstable
- but, can have high reliability and not measure what intended to
types of correlational research
cross-sectional: take snapshot of many variables at 1 time
longitudinal: measuring variables repeatedly at dif time points
correlation good with ecological validity, not causation
Humes requirements
- contiguity: cause & effect occur close together in time and space
- precedence: C before E
- necessity of cause: if E, C must have occurred
Mill’s addition - sufficiency of cause: if C, E must have occurred
confounds
when causation is unable to be determined bc of measurement
*tertium quid: a third, unmeasured variable
Eliminating confounds experimental research *random assignment *counterbalancing but, procedural confounds cannot be controlled
types of variation
systematic: effect of condition, any confounds
BW groups unsystematic: difs bw individuals, measurement errod
W/IN groups unsystematic: instability of difs bw individuals, measurement error
W/in versus b/w groups design
W/in
- has greater power because on most measures, ppl tend to be stable and substantial
- therefore, they have lower unsystematic variation, so all thats left is condition effect
w/in groups confounds
- passage of time: change in group bc of time passed. (bored, hungry)
- carryover effects: changes due to taking 1st condition: practice effects, lost confidence, bored)
solutions:
*counterbalance test order
but, this may increase unsystematic variance and reduce power. also some intervention needs order
how distribution can deviate from normality
skew: lack of symmetry
* positive- pile on left
* negative: pile on right
kurtosis: pointyness
* lepto: positive, pointy, many scores in tails
* platy: negative, flatter, thin in tails
* meso: score of 0, normal
measures of dispersion
range, IQR, SS,
variance: avg dispersion
standard dev: how scores deviate from mean
Z-scores
a way to standardize raw scores so that they are all on same comparison scale
used in probability distributions
mean=0, SD=1
if z dist is normal = standardized normal distribution
