Behavioral Science/Stats Flashcards
Study type that establishes prevalence
Cross-sectional
Case-control study measure?
Odds ratio (What was “odd” about their history?)
Case-control study design?
People with disease v. people without disease. Looking for prior exposure or a RF
Cohort study measure?
Relative risk (Increased risk for disease)
Cohort study design?
People with exposure v. people without exposure
Twin concordance study design?
Frequency of disease between twins
Pharm phase I asks?
Is it safe? Safety, toxicity, pharmacokinetics
Pharm phase II asks?
Does it work? Efficacy, dosing, AE’s
Pharm phase III asks?
Is it as good or better?
Pharm phase IV asks?
Can it stay? Post-marketing surveillance
Sensitivity =
TP / (TP + FN) = Pr (T+|D+)
Specificity =
TN / (TN + FP) = Pr (T-|D-)
PPV =
Pr (D+|T+) = TP / (TP + FP)
NPV =
Pr (D-|T-) = TN / (TN + FN)
Prevalence is related to incidence how?
Prevalence is approx incidence rate x avg disease duration
Incidence rate =
of new cases in specified time period / population at risk during same time period
Odds and Probability related how?
Odds = Pr / (1-Pr) ; Pr = O / (1 + O)
What parameters are only dependent on the test itself?
Sensitivity, specificity, likelihood ratio
Odds ratio and relative risk?
OR = (a/b) / (c/d) = ad/bc; RR = [a / (a+b)] / [c (c+d)]
RRR =
1 - RR
Difference between effect modification and confounding bias?
Effect modification is something to be studied; Present when effect of the main exposure is modified by presence of another variable.
Power
= 1 - Beta; Beta = Pr(type II error) = Pr of rejecting null hypothesis when it is truly false = Pr difference when there IS one = Maximize for clinical trials!
Type 1 v. type 2 error
1st and foremost, making something out to be when there isn’t anything. (Rejecting H0 when it is true). 2nd, saying nothing when there is (Fail to reject, when H0 is false)
Berkson’s bias
Selection bias - selecting hospitalized patients as the control