ebp study design Flashcards
experimental design
include active manipulations
observational design
no manipulations. observe systematically, don’t alter
controlled experiemental
include non-treatment comparison group
-controlled trial: one group receives treatmnt/manipulation, control group does not
uncontrolled experimental
no control group: all participants receive treatmemt (no control/comparison group)
ex. initial drug safety, tolerability determination
prospective
hypothesis testing, methods planned out before data collection
-experimental studies must be prospective
retrospective
analyze pre-existing data
-ranked lower than prospective: no control over systematic or unknown influences, can’t assess validity of procedures
*key: you form your question with the data that already exists
representative sample
-even representation from all relevant group (gender, ethnicity, age, region)
-stratified sampling is great for this
matched groups
purposefully balance groups on traits expected to affect outcomes
longitudinal
following the same people over a period of time
limitations: subject to atrition, costly, have to wait for results
cross-sectional
use subject groups of different ages/stages to proxy for developmental trajectory
strengths: can measure just once: no subject attrition, cheaper, likely to finish data collection
limitations: risk nonequivalence of important/nuisance variables between groups
case studies
an observational control study, describes single patient
case series - describe series of similar patients
pretest/post test designs
pretest: pretest - treatment/alternative - test
posttest: treatment/alternative - test
correlational study
looking for a relationship among data that already exists
always retropsective
internal validity
accuracy of relation between observations and the subjects observed
-does it measure what the authors intended it to
external validiy
generalizability = applicability of patterns/results to a larger population
-does it apply to more people
blinding
make involved people unaware of information that could bias findings
examples:
-researchers dont know which group is which
-participants are blind to what group they are apart of
-both are blinded, people doing analysis are blinded
confounders/nuisance variables
unintended, uncontrolled, or unknown factors that could affect the results (alternate explanation, nullification, or false conclusion)
-mitigate by knowing behavior might affect and try to eliminate it
replicability
do the studies replicate
descriptive stats
summarizes characteristics of data set
-counts: frequency, percentage
-location/central tedency: mean, median, and mode
-indvidual location: rank, percentile rank, standard score
-variability: range, variance, standard deviation
inferential stats
use sample to infer characteristics of population (t-test, anova)
p-values
if p=.05, 5% of the time you’d find a difference just by chance, even when there isnt one
p-value < alpha value = reject null hypothesis/accept alternative hypothesis
confidence intervals
range of values around a descriptive stat of our sample (eg. mean, medium) that were x% confident contains the populations true stat
-typically 95%
parametric
assume (=require) data have normal distrubution
-two sample t-test
-paired t-test
-simple or complex anova
-pearson correlation (r)
non-parametric
dont assume/require anything an=bout the shape of the distrubution
-wilcoxon rank-sum test
-wilxcoxin signed-rank test
-kruskall wallis test
-friedmans test
-chi-square
-spearman correlation (p)
