Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
50 Cards in this Set
- Front
- Back
research strategy
|
general plan of attack
|
|
research design
|
specific plan and structure of the investigation. development of a plan for selecting and measuring IVs and DVs in order to answer research questions about their relationships (the purpose of the design)
|
|
controlling variance in the research design
|
maximize systematic variance associated with IV and DV, and minimize random error or random variance including error of measurement. control extraneous variables' influences on DV
|
|
difference in purpose between experimental and descriptive research
|
experimental: cause and effect. descriptive: differences between groups as they exist in their natural habitat. both trying to answer the research question empirically and reduce contamination by extraneous variables
|
|
group designs
|
1 or more groups exposed to 1 or more levels of IV and average performance of group on DV is examined. applies to both experimental and descriptive.
|
|
single subject design
|
individual performance, but perhaps more than one person in experiment (but each is still compared to themselves). multiple subjects being looked at independently in the study. performance of individual on DV measured. usually multiple measures, time series. baseline segment--> treatment segment (AB).
|
|
internal validity
|
extent to which question is answered with minimal extraneous variance. have to control for ex variables, random variability, measurement error. too controlled=not generalizable. has to be assured before ex validity.
|
|
external validity
|
generalizability to other groups, important when doing treatment efficacy research. trying to generalize results from uncontrolled experiments is pointless. difficult to establish from a single study.
|
|
experimental: between subjects designs
|
two levels of IV (bivalent), manipulation occurring. looking at differences between a group. performance on DV is measured, and the differences is called the INDEX OF EFFECT. randomization and matching to control subject selection and achieve group equivalence.
|
|
randomization
|
everyone has an equal chance of being in any group. variables to be controlled are distributed randomly. statistical randomization: influence is constant, the larger the "n" the more likely the occurrence
|
|
matching
|
matching up on all variables except for the one being studied (match aphasics with non-aphasics on age, gender, and edu level). 2 groups matching in freq of extraneous variables. matching pairs (individuals) and randomization are better for smaller "n"s
|
|
descriptive between subject designs
|
may be used with comparative, cross sectional, and survey. multi or bivalent. cannot randomly assign. equate subjects on extraneous variables
|
|
within subject designs
|
performance of same subject compared to different conditions. at all levels of IV, no randomization of subjects. caution: sequencing or order effects, learning effect, interference. decrease these by randomization and counterbalancing
|
|
mixed designs
|
more than one IV, one between-subject and one within-subject. may combine descriptive and experimental methods. used when one group performs poorly, look to see what predicts behavior.
|
|
characteristics of baseline behavior
|
level=overall value, trend=shape of segment (flat, increasing, decreasing), slope=rate of change over time, variability=range of behavioral fluctuation (error, prefer <5% baseline)
|
|
red flags for internal validity:history
|
what happens between measurements (long time gap in longitudinal studies). life gets in the way
|
|
red flags for internal validity: maturation
|
what changes within the subject during the course of the study? changes due to normal development. is the change due to intervention or to normal development?
|
|
red flags for internal validity: testing or practice effects
|
what effect does taking one test have on taking the next test. did one subtest have clues for another subtest? need to counterbalance the order of subtests.
|
|
red flags for internal validity: instrumentation
|
what changes are there in calibration of the instruments?
|
|
red flags for internal validity: statistical regression
|
what change in performance is there in good/bad subjects? what do they do with the outliers?
|
|
red flags for internal validity: differential selection of subjects
|
what differences exist between groups not related to the study? is there something about those subjects that might have an effect?
|
|
red flags for internal validity: mortality
|
how many subjects are lost during the course of the study?
|
|
red flags for internal validity: interaction of factors
|
how many of the threats to internal validity co occur?
|
|
red flags for external validity: subject selection
|
do the subjects selected represent the whole?
|
|
red flags for external validity: reactive or interactive effects of pretesting
|
do subjects change just by going through pre-testing?
|
|
red flags for external validity: reactive arrangements
|
does the testing site itself determine the performance? using sound proof booths, etc. not very generalizable
|
|
red flags for external validity: multiple treatment interference
|
does the fact that there are multiple treatments change the performance on each individual treatment?
|
|
improving external validity: random sampling
|
all subjects in study have equal probability of being selected. not always practical, so sample from cluster from specific subpopulation or sample from stratified groups of subjects
|
|
improving external validity: increase sample size
|
use power analysis or effect size. power analysis determines the probability of detecting a hypothesized degree of effect in a sample size. if the effect size is sufficient, theres a good probability of detecting the effect in the sample size used. the more variable the data, the more participants needed.
|
|
improving external validity: replication
|
a study that can be repeated and produce the same results over and over is stronger. direct rep: same subjects/different researcher. systematic rep: variation in subjects or treatment.
|
|
statistics
|
computed estimates of parameters. numeric descriptors of sample population. aids in answering research question by indicating how plausible certain conclusions are in light of the obtained data. statistical regression: always going to have people performing at the extremes. did the researcher base his research on the extremes?
|
|
nominal level of measurement
|
naming, subjects are not put in an particular order. no category is higher or lower. ex: marital status, employment status, residence.
|
|
ordinal level of measurement
|
labels placed in order from high to low. ranks do not indiciate the amount of difference between subjects. ex: scale 1-5. not any specific distance between labels
|
|
interval scale level of measurement
|
equal intervals between ranks, no absolute zero, ex: test score. all points equal.
|
|
ratio scale level of measurement
|
equal distance between scores (like interval), has absolute zero rank, for ex: weight, time
|
|
measurement error
|
score depends on true score (that person's true ability) + measurement error (variations). error caused by factors that distort the true score (iq testing, freq counts, transient and stable attributes, situational factors, characteristics of the measure itself, actual mistakes)
|
|
reliability
|
the degree to which a measure is consistent in producing the same rating when measuring the same things. consistently show differences between individuals who are truly different. 0=no reliability/1=perfect reliability. neither occur in the real world.
|
|
measures of reliability: internal consistency
|
(coefficient alpha), a reliability measure that shows internal consistency. degree to which scores on items in a scale correlate with each other.
|
|
split half correlations (reliability)
|
correlating half of the items with the other half
|
|
test-retest correlation
|
the correlation between scale scores obtained on a test at one time and score on the same scale taken at a different time.
|
|
inter-rater reliability
|
the ratio of agreements between two or more observers or raters to the total number of observations
|
|
parallel forms
|
correlation between the scores of two closely related forms of the same scale (two tests on vocab)
|
|
face validity
|
degree to which a measure appears to assess what it is supposed to assess.
|
|
content validity
|
degree to which test items correspond to the content of the assessment
|
|
criterion validity
|
degree to which test correlates to another outside validating criterion (ex: iq test)
|
|
predictive validity
|
degree to which scores on a test predict later behaviors or scores
|
|
concurrent validity
|
degree to which test correlates with another conceptually related test
|
|
construct validity
|
degree to which a test has a pattern of correlation with other variables that would be predicted by a sound theory
|
|
method
|
structural framework of article. describes the participants, the materials employed, how the materials are used, the procedures. identify the research strategy
|
|
experimenter bias
|
experimenter attributes interfere to influence behavior of participant, expectancies change how observer rates behavior. can be controlled by blind technique (experimenter doesnt know which is control and which is experimental), or by investigator and experimenter being different people. effects validity.
|