predictive_value() computes the positive or negative
predictive value of a diagnostic test, showing sensitivity, specificity
and prevalence as separate steps. Supplying prevalence
applies the test to a population other than the one in the table, via
Bayes’ theorem, which is how a test validated in a clinic is assessed as
a population screen.confint() no longer silently ignores its
level argument. Intervals are computed when the measure is
computed, because the derivation displays the arithmetic that produced
them, so level still cannot change the result – but it now
warns and names the fix (conf_level on the measure) instead
of returning an interval at a different level than the one asked
for.
Step labels no longer paste a custom exposure label directly
after “the”, which produced text such as “Odds of Ill among the ate
potato salad”. Labels now name the epidemiological role and give the
user’s label in parentheses: “Odds of Ill among the exposed (Ate potato
salad)”. Output is unchanged when the default exposure
labels are used.
First release.
epi2x2() builds a 2 x 2 table in standard
epidemiological orientation.risk_ratio(), odds_ratio() and
risk_difference() return the measure together with the full
worked derivation.attributable_fraction() computes the fraction among the
exposed or in the population, showing that the direct, Levin and
Miettinen formulas agree.epi_strata() and collapse_strata() handle
stratified data.mh_odds_ratio() and mh_risk_ratio() pool
across strata, printing the stratum weights and the crude estimate
alongside the adjusted one.homogeneity() implements the Breslow-Day test with
Tarone’s correction.check_work() compares a hand calculation against every
step of a derivation and reports where it diverged.steps_table() returns a derivation as a data
frame.options(epibyhand.verbose = ) and
options(epibyhand.digits = ).