research design Flashcards
research design is the
overall plan for answering the research question
question should have
- nature of comparison
- type of setting
- population, sample
- independent and dependent variable
comparisons between two groups or more
- single group: 2 or more points in time (pre/post)
- single group under different circumstances or experiences (group vs individual therapy)
- based on relative rankings (severe vs mild autism)
- compare with samples from other studies
research designs also include
- methods to be used to contol variables (isolate dependent and independent variables)
- timing and frequency of data collection (when, relative to other events)
- setting (naturalistic vs laboratory)
- nature of communications with subjects (fully divulge or not)
dimensions along which designs can be described
- experimental vs non-experimental
- degree of structure imposed
- time dimension
- type of group comparisons between subjects and within subjects
- non-experimental
experimental
quantitative
non-experimental
qualitative
degree of structure imposed
how much we are controlling
time dimension
- longitudinal
- cross-sectional
longitudinal
looking for change in time
cross-sectional
collecting data at 2 points in time
multiple points of data collection secondary to
studying time related processes
determine time sequences
developing comparisons
enhancing research control
trend studies
periods of time
observing what’s going on with trends
cohort
when you take a small population (by age)
study them over time with respect to a phenomenon (subject)
panel studies
take same cohort and measure across 2 periods of time
follow up studies
how they feel about same topic in a period of time
ex 5 years down the line
cross section cohort
comparing 2 different cohorts follow up study can be a panel study
within subjects
in same group of people (more control of variables)
between subjects
different group of people
non-experimental
not manipulating anything (qualitative)
- retrospective
- prospective
retrospective
have an outcome
want to know what causes are (antecedent) study prior habits
prospective
know causing variables
going forward in time
see how many develop outcomes
usually do prospective after you gain evidence from
retrospective
cons of retrospective
memory might not be accurate
might get defensive
might not tell the truth
characteristics of a good design
- must be appropriate to question asked
- minimize biases that can distort results of study
- precision must be enhanced
- power should be adequately dealt with
biases result from
- differences among participants in groups
- researchers preconceptions
- use triangulation
differences among participants in groups
more so in non-random groups
researcher’s preconceptions minimize this
can do a double blind study
neither participant nor data collector knows specific research objectives
use triangulation to
reduce bias
at least 3 points of data collection
precision must be enhanced
sensitivity with which effect of independent variable relative to effects of extraneous variable (confounding) can be detected
want it high
power should be adequately dealt with
ability of design to create maximal contrast amongst group being compared
one can detect relationship between variables
experimental designs
characteristics of true experience
characteristics of true experience
- manipulation
- control
- randomization
manipulation
independent varibale
control
not getting any treatment
outside things that impact treatment
ex control environment
randomization
pick random from different population extreme view will be cancelled equal chance that they can be picked for any group -table of random #s -cluster randomization -matching
table of random #’s
flipping a coin, most reliable method for equating groups on all possible characteristics that could effect outcome of study
cluster
picking people in clusters, random, pick a bunch to represent zip code
matching
people in each group match and have equal numbers
experimental designs
- basic
- Solomon 4 group design
- factorial design
- repeated measures
- randomized control trials
posttest only
someone’s been given a treatment and measuring outcome after the fact
pretest
measuring outcomes before and after
basic
- posttest ony
- pretest/postest
Solomon 4 group design
take into account influence of pretesting on subsequent posttest results
factorial design
1 independent and 1 dependent
- main effects
- interaction effects
main effects
1 variable
interaction effects
combine independent varibales
quasi experimental designs
- manipulation + control or randomization
- non-equivalent control group design
- time series designs
- times series: non-equivalent control group design
non-experimental designs 2 broad classes
- ex-post factor research
2. descriptive
ex-post factor research
coorelational
descriptive
- descriptive correlational
- univariate descriptive
- retrospective
- prospective
other non-experimental types
- survey research
- evaluation research
- needs assessment
- secondary analysis
- meta analysis
- historical research
- case studies (single subject designs)
qualitative designs field studies
- ethnography
- phenomenology
- grounded theory
ethnography
roots in anthropology
phenomoneology
roots in philosophy
ethnomethodology
roots in sociology
ethnomethodology
roots in sociology
strategies in a research study
- triangulation
- mixed methods
- fully integrated
integrated research designs 2 broad categories in multi-method research
- component
- integrated
component
- triangulated
- complementarity
- expansion
integrated
- iterative
- embedded or nested
- holistic
- transformative
advantages of experimental designs
most powerful method available for testing hypotheses of cause and effect relationships between variable
disadvantages of experimental designs
many situations in which experimental design is impossible, secondary to ethical or protocol considerations, artificiality or setting, randomness not always real
Hawthorne effect (placebo effect)
knowledge of being included in a study may be sufficient to cause people to change their behavior
study could have double hawthorn effect
when those conducting study also change their behavior
therefore, do a double blind study so neither researcher nor subject know
repeated measures design
-one group tested under all conditions (within subjects design)
repeated measures design
-one group tested under all conditions (within subjects design)
one group tested under all conditions (within subjects design) strong design because of
ability to control potential influence of individual differences
one group tested under all conditions (within subjects design disadvantage is practice effects
exposure to test
one way to combat this is increase length of time between treatments, so subject’s scores can go back to baseline
one way repeated measures design
one group of subjects exposed to all levels of one Independent variable
order effects can occur, that is potential biasing effect of test sequence, so randomize order of presentation for each subject
one way ANOVA test
crossover design
- only 2 levels of individual variable
- half receive treatment A, then treatment B
- half revieve treatment B followed by treatment A
- again add time in between to remove any testing effect or treatment residual effect
- 2 way ANOVA, paired t-test
two way design with two repeated measures
- 2 individual variable with two levels each
- subjects get all 4 conditions
- 2 way ANOVA
efficacy
examining benefit of treatment as compared to control group or standard of care group
effectiveness
benefits and use of procedures under real world conditions
quasi experimental designs
- manipulation + control OR randomization
- non-equivalent control group design
- time series designs
time series
non-equivalent control group design
quasi experimental designs: one group-pre-test/post-test design
paired t-test, on small samples with ordinal data-Wilcoxon sign rank
quasi experimental designs: one way repeated measures design
one way ANOVA
quasi experimental designs: time series design
- multiple measures before and after tx, to document trends
- use graphic visual analysis
quasi experimental designs: non-equivalent pre-test/post-test control group design
- groups are not equal
- paired t-test, ANOVA
quasi experimental designs: non-equivalent post-test only control group design
- regression analysis
- discriminant analysis
non-experimental designs: 2 broad classes
- ex-post facto research
- descriptive
non-experimental designs: ex-post facto research
- after fact (intervention)
- correlational
correlational
- degree of association among variables
- it is a function of covariation in data
non-experimental designs: descriptive
- descriprive correlational
- univariate descriptive
- retrospective
- prospective
descriptive coorelational
describe nature of existing relationships
univariate descriptive
one variable
retrospective
examine data collected in past
prospective
current data and follow up in future
non-experimental designs (quantitative)
- survey research
- evaluation research
- needs assessment
- secondary analysis
- meta analysis
- content analysis
- historical research
- case studies
survery research
statistical
evaluation research
- how well something is doing? (standardized)
- what is going on?
needs assessment
collect data on what place is like/what they need may have multiple needs
secondary analysis
take existing data set ask different research question and alalyze
meta analysis
take a bunch of study that are asking same question and their population combine statistically (quantitative)
content analaysis
- can be numerical/qualitative
- frequency of idea
historical research
go back and look at old records/artifacts to recreate history
cases studies
look at individual in depth every client factor look at them historically
-single subject designs