Behaviorial Science/Biostats Flashcards
power of a study increases w/
increased sample size, expected effect size, precision management
power =
1 - beta
beta = stating there ISN’T an effect/difference when ONE EXISTS
sensitivity =
test detects dz when dz present; screening test (rule out a dz)
1 - FN rate
TP/(TP + FN) aka TP/total dz +
a/(a+c)
specificity =
test indicates NO DZ when dz actually absent; confirmatory test (rule in a dz)
1 - FP rate
TN/(TN+FP) aka TN/total dz negatives
d/(b+d)
PPV
person WITH dz tested positive in test
TRUE POSITIVE!!
TP/(TP+FP)
a/(a+b)
high pretest probability = high PPV
NPV
person WITHOUT dz who tested negative in test
TRUE NEGATIVE!!
TN/(FN+TN)
d/(c+d)
high pretest probability = LOW NPV
incidence
of new cases
prevalence
incidence x time
looks at ALL current cases –> tx that prolongs survival INCREASES prevalence
odds ratio (OR)
used in case-control studies
odds that group w/ dz exposed to risk factor divided by odds that group w/o dz exposed to risk factor = (a/c)/(b/d) =
ad/bc
relative risk (RR)
used in cohort studies
risk of developing dz in exposed group divided by risk in unexposed group
[a/(a+b)]/[c/(c+d)]
if low prevalence –> RR = OR
attributable risk
difference in risk b/w exposed/unexposed groups
Attributable Risk % = (RR - 1)/RR
[a/(a+b)] - [c/(c+d)]
event rate(of tx) - event rate(in other group)
Absolute risk reduction (ARR)
AR(in control) - AR(w/ tx)
[c/(c+d)] - [a/(a+b)]
needed to treat
pts needed to treat for 1 pt to benefit
1/ARR
needed to harm
of pts who need to be exposed to RF for 1 pt to be harmed
1/attributable risk
relative risk reduction (RRR)
[AR(control) - AR(tx)]/AR(control)
type 1 error (alpha)
stating there IS an effect when there ACTUALLY IS NOT (mistakenly reject null hypothesis)
b/(b+d)
type II error (beta)
stating there IS NOT an effect when there ACTUALLY IS (fail to reject null hypothesis)
c/(a+c)
95% confidence interval Z-score =
1.96
99% confidence interval Z-score =
2.58
standard error of mean (SEM) =
SD/(sq rt of n)
SD (standard deviation) = aka sigma
n = sample size
formula to obtain confidence limits
mean +/- [Z-score x SEM]
wave seen in awake stage?
beta (highest freq, lowest amp)
wave seen in relaxation stage?
alpha
wave seen in N1 stage?
theta
