Research Basics Flashcards
Independent Variable
-manipulated
-intervention
Dependent Variable
-measured
-change in strength
Confounding Variable
-variable that could influence outcome of the study
Quantitative Research
-uses numbers
-IV manipulated
-reduce confounding variable
-t-test, ANOVA, mean med mode
Qualitative Research
-understand a problem from the perfpective of the affected population
-interviews
Strengths:
-descriptions
-human side of issues
Limitations:
-time and cost
-cannot measure validity and reliability
-cannot be generalized
Single-Subject
-one or few participants measured several times
-usually unique pop or intervention
Nominal Scale
-qualitative
-identified only by name
-show differences in individuals
ex: gender, disease, zip code
Ordinal Scale
-qualitative
-ordered categories
-direction of differences between individuals
-no true number value
ex: no help, some help, independent, MMT
Interval/Ratio Scale
-quantitative
-ordered series of equal size
-direction and magnitude
Interval: zero is arbitrary
Ratio: real zero
ex: feet, temp, ROM, speed
Descriptive Studies
-Retrospective
-Normative
-qualitative
-describes data
Exploratory Studies
-correlational: relationships
-predictive: reliability and validity
-case control (quasi experimental)
Experimental Studies
-RCT only true
-cause and effect
Quasi-Experimental
-no manipulated IV
-pre exisiting variable
-cohort studies
Alpha Level
-predetermed before study
-pre set significance level
-0.05 usually, 5% chance data relationships are not significant
Smaller: high risk
Larger: important even if theres a chance at not being effective
P-Value
-determined by outcome of study
-actualy probability that results occured by chance
- <0.05 to be significant
Validity
-info is believable and useful
Internal Validity of Study
-did the IV cause the change in DV
-RCT is the best design to maximize
Internal Validity: History
-something happend between pre and post to change result
Internal Validity: Maturation
-did participants change over time
ex: children aging, disease progression
Internal Validity: Attrition/Mortality
-who dropped out and why
Internal Validity: Repeated Testing
-did the repetiitons change the outcome
ex: doing the same thing for weeks will ensure better performance
Internal Validity: Instrumentation
-was the instrument changed
-calibrated
Internal Validity: Regression to the Mean
-groups with extreme scores tend to regress towards mean
ex: really bad pt will get better and really good might stay or get worse closer to the mean
Internal Validity: Experiementer Bias
-biased
Internal Validity: Selection
-did the groups differ
-sample not representative of population
Contruct Validity
-are we measuing what we think we are
External Validity for Study
-can the results be generalized by the population
-is it specific enough to have a difference but not too specific that you cant generalize it
Statistical Conclusion Validity
Low Power: small sample size, too much variability
-sample size too small to be representative
-used the wrong test
-error rate (Type 1: reject null when shouldnt or Type 2: fail to reject when should have
Simple Random Sample
-everyone has the same chance
Systematic Sampling
-select in a certain order
ex: every 5
Stratified Sampling
-select sample then divide into parts
-divide a population into individuals
ex: 10 students from every age group
Cluster Sampling
-double random
-random population and randomly divide into groups
ex: 5 random schools and select 10 random people from each of those schools
Convenience Sampling
-members volunteer or self select
-usual
Sampling Error
-sample doesn’t match population
Cohen’s D
-measure of effect size: meaningful amount of change
-difference in means divided by pooled SD
-generally larger with bigger differences
less variability = larger ES
d= change in means/ poooled SD
Small= 0.2
Medium= 0.5
Large= 0.8
Power
-chance of finding a sig diff
-affect by effect size and participants
Validity
-how correct it is
-measures what it should
-cannot exit without reliability
Reliabiliy
-how consistent it is
-degree of association
-can exist without validity
Contingency Table
True Positive: a; tested positive with test; have the condition
False Positive: b; tested positive; don’t have the condition
False Negative: c; tested negative; have the condition
True Negative: d; tested negative; don’t have condition
Sensitivity
-snout: rule out
-true positive test
-shows all the positive so it rules out the negatives
-a/(a+c)
Specificity
-spin: rule in
-true negative
-shows all the negative so it rules in the positives
-d/(b+d)
Predictive Value (+)
-likelyhood that the positive test = having condition
-a/(a+b)
Predictive Value (-)
-likelyhood that the negative test= not having condition
-d/(c+d)
Positive Likelihood Ratio
-increased odds of having condition if testing positive
-ratio of true to false
-Sensitivity/ (1- Specificity)
-higher= more likely
Negative Likelihood Ratio
-decreased odds of having condition if testing negative
-ratio of false to true
-(1- Sensitivity)/ Specificity
-lower= less likely
Guide to Interpreting LR
-most powerful tool for quantifying importance of a particular test
10+, Best, inc by 45%
5+: OK, inc by 30%
2+: Poor, inc by 15%
1.0-: Useless, 0%
0.5-: Poor, dec by 15%
0.2-: Ok, dec by 30%
0.1-: Best, Dec by 45%
Minimal Detectable Change
-MDC
-amount of change needed to overcome measurement error
-increase reliability of test decreases MDC
Minimal Clinical Important Difference
-MCID
-amount of important change from the perspective of individual
-should be bigger than MDC
One-Way Repeated Research Design
-one group doing the same thing over titme
Posttest-Only Randomized Group
-randomized
-2+ groups that are only measured after the intervention
Posttest-Only Non-Randomized Research Design
-non randomized
-2+ groups that are only measured after the intervention
Factorial-Fully Independent Research Design
-2+ interventions at the same time
-each person stays in the same group the whole time
Factorial-Fully Repeated Research Design
-2+ interventions at the same time
-all participants switch groups throughout the study
Factorial Mixed/Split Plot Research Design
-with or without randomization
-Participants in designated groups move through several parts
-most common in PT
Crossover Research Design
-with or without randomization
-groups switch interventions after 1st is done
Measurement Theory
-psychometrics
-foundation for evaluating tests and their uses
-reliability and validity are most fuindamental measurment theory
Measurement Study
-any study that investigates the reliability or validity of a research measure
Operational Definition
-objective variables must be defined in study
Inter-Tester Reliability
-different raters get the same score
-tester reliability
Intra-Tester Reliability
-same rater coninuously gets the same score
Test-Retest reliability
-is the instrument consistent enough to get the same score
-is the patient consistent enough to get the same score
-instrument reliability
Agreement
-if 2 ratings are similar or match exactly
Parallel Reliability
-aka equivalent
-2 of the same tests are given to 2 groups
-do the groups meausre similarly
Split-Half Reliability
-questions from the same sources are both giiven to one group
-does the group measure the same on each
Face Validity
-does it measure what it’s supposed to
Content Validity
-dooes it measure the entirety of what it is suposed to according to experts
Criterion-Based Validity
-degree to whih the outcomes correlate with the gold standard
Concurrent Validity
-degree to which the outcomes correlate with another test (gold standard or not)
-given at the same time
Predictive Validity
-can it be used to predict some outcome
-Berg balance
Construct Validity
-degree to which a theorhetical construct is measured by an instrument
Responsiveness to Change
-extent to which significant changes in the participants are reflected
-ceiling/floor effects alter this
-more items: more responsiveness
ROC (Receiver Operating) Curves
-Sensitivity vs False positives
-Increased area under the line= high sensitivity and specificity
-want greater than 1/2 under the line