Hoofdstuk 14: measurement and data quality Flashcards
what is measurement in quantitative research ?
=structured measurement of constructs
- biophysical
- self-report
- observational
What is the error of measurement ? and sum up some sources of bias
–> Obtained score= true score +- error
sources:
- transient personal factors
- situational contamination
- response-set bias
- administration variations
- instrument clarity
- item sampling: wich items are included
What are the major types of measure ? (2)
Reflective scales: Construct forms the items: eg symptoms caused by depression
Formative scales: Items form construct: eg score on stressful life events defines susceptibility to stress induced illness
Define the cosmin polit yang taxonomy for scores(cross sectional) (2) (each 3)
- Reliability:
- reliability
- internal consistency
- measurement error - Validity
- Content and face validity
- Criterion validity
- Contruct validity
Define the cosmin polit yang taxonomy for change scores(Longitudinal) (2) (1 each)
- Reliability
- Change score reliability - validity
- Responsiveness
In the reliability domain explain reliability (+4)
=degree to wich scores are the same for repeated measures (consistency)
- Test retest reliability
- Interrater reliability
- Intrarater reliability
- Parallel test reliability
In the reliability domain explain internal consistency
=degree to wich items measure the same construct
statistic: cronback alpha–> higher the better
In the validity domain explain face and content validity
a) Face: does it look like its measing what we want to measure ? (niet wetenschappelijk)
b) Content validity: does the instrument cover everything ? (relevence, comprehenceveness, balance) –> through expert panels
In the validity domain explain Criterion validity
=is the measure consistent with a criterion ? (gold standard)
- concurrent: is your measure the same as a measure that looks like yours ?
- predictive: does your measure correlate with a criterion measured in the future ? (eerdere test voorspeld uitkomst)
describe the diffrent relationships between variables (dichotomous, continous) in criterion validity
-continuous measure + continuous criterion: pearsons r (eg 2 depression scales)
-dichotomous measure + dichotomous criterion:
=>diagnostic acuracy
a) sensitivity –> rate of true positives
b)specificty –> rate of true negatives
=>predictive value
a)positive pred. val.: proportion of positive results who have the criterion
b)negative pred val.: proportion of negative who do not have criterion
=>likelyhood ratio:
a)LR+: true positive/false positive=sensitivity/(1-specificity)
b)LR-: false negative/true negative=(1-sensitivity)/specificity
-continuous measure + dichotomous criterion
=> ROC-curve –> best balance between sensitivity and specificity
-> AUC: index of criterion validity close to 1 is desirable
In the validity domain explain constuct validity (3)
=what attribute is really being measured ?
- Hypothesis tesing construct validity
a) convergent: construct X &Y are related; A measures X and B measures Y
- -> A and B correlate ? ==> V:A is a valid measure of X
b) discriminative(known groups): two or more groups differ relating to construct
c) divergent(discriminant): both measures dont measure same construct ?
- -> post partum depression scale and a maternal anxiety scale = divergent - structural validity: what dinemsions of the construct are captured by the instrument ? and are these consistent with theory ?==> factor analysis
- Cross-cultural validity: translated or adapted instrument equivalent to orignal ?
In the change scores (longitudinal) describe the reliability(1) and responsiveness(1)
-Change score RELIABILITY: amount of change between two measures (!measurement error)
Responsiveness: =validity of change scores:
