Kraushar Flashcards
2 X 2 tables for hypothesis testing:
(alpha, beta, and power)
Cumulative Incidence Definition and Equation:
(new cases) / (total population at risk over time)
- fraction of people initially free of the outcome but who develop it over a period of time
Null Hypothesis (Ho):
- states that there is no difference
Absolute risk difference =
Iexposed - Iunexposed
- incidence in exposed group minus incidence in unexposed group
Incidence equation in steady state:
prevalence / duration
Relative risk =
Iexposed / Iunexposed
- incidence in exposed group divided by incidence in unexposed group
Odds ratio definition and table:
- retrospective
- start with people who have disease, and go backwards to find risk factor
Duration equation in steady state:
prevalence / incidence
Prevalence equation in steady state:
(incidence) X (average duration)
2 X 2 tables for hypothesis testing:
(type 1 and 2 errors)
The probability of two mutually exclusive events, A or B, occurring =
(probability A) + (probability B)
Type II error:
- FALSE NEGATIVE
- saying there is no difference in treatment effects when there is.
- failing to reject (accepting) the null hypothesis when it should be rejected
power =
1 - beta
- power of study to pick up a difference when it actually does exist
Incidence Rate definition and equation:
(new cases) / (total time lapsed)
- rate at which new disease has occurred in the population at risk per some unit time
Normal distribution:
- classic bell curve
- highest density in middle, tapers off on both sides
- mean, median, and mode all in the same place (dead center of bell curve)
- dead center for all these = perfect normal distribution (Gaussian)
2 X 2 tables for hypothesis testing:
(false hits and false misses)
Type I error:
- FALSE POSITIVE
- saying there is a difference when there is not.
- rejecting the null hypothesis when it should be accepted
Prevalence Definition and Equation:
(total cases) / (total population)
- fraction of people experiencing a condition at a given point in time.
- number cases right now
Relative risk definition and table:
- prospective; follow
- start with the risk factor, follow them over time, see how many people get disease
- everyone starts with no disease
Probability:
- a proportion in which the frequency of both events are in the denominator:
A/(A+B)
alpha =
- the probability you’ll make a false hit.
- Arbitrarily, p = 0.05
- false hit = type 1 error
- 5% of the time, we’ll make an error
Power depends on:
- sample size (better power = more participants)
- size of effect (better power = stronger effect)
- subject compliance
Sensitivity equation and table:
TP/(TP+FN)
SNOUT
Standard deviation:
+/- 3SD contains –% of observations
99.7%
beta =
- probability you won’t find a difference when one actually exists (false miss)
- false miss = type 2 error
- probability of missing a reality
A specific test should be utilized when:
- false-positive can harm the patient physically, emotionally, or financially.
- used to “rule-in” diagnoses when data suggest.
Mean, median, and mode describe:
“central tendency”
Aboslute risk =
incidence
Standard deviation:
+/- 1SD contains –% of observations
68%
Bimodal distribution:
- think of breasts - there is a variable (like sex; M/F) under the bimodal curve
- highest density at both bells, tapers off in each direction evenly
- bimodal has two “density centers”
The probability of two independent events, A and B, occurring together =
(probability A) X (probability B)
Standard deviation:
+/- 2SD contains –% of observations
95%
A sensitive test should be chosen when:
- there is an important penalty for missing the diagnosis.
mean, median, and mode location in a normal distribution:
overlapping dead center of the bell
Specificity equation and table:
TN/(TN+FP)
SPIN
Mean, median, and mode regarding skew:
- mode insensitive (stays in bell)
- median moderately sensitive
- mean very sensitive (goes out to far tail)
Right skewed distribution:
- tail/outliers on the right of the bell; RIGHT TAIL
- highest density on the left, tapers out toward the tail
- Mode is where the bell peaks; mean is far in the tail. Median in middle.
Left skewed distribution:
- tail/outliers on the left of the bell; LEFT TAIL
- highest density on the right (where the bell is), tapers out toward the tail
Odds:
- a ratio of the frequency of:
A to B, or A:B or A/B
Absolute risk definition and table:
- attributable risk
- a difference of relative risks
- how much of getting the disease is attributable to having the risk factor
The three main criteria for abnormality:
- unusual
- associated with disease
- treatment does more harm than good
