An internal function called by the tmle function to calculate the population mean effect when there is missingness in the data, but no treatment assignment. When observations are in treatment and control groups, estimates the additive treatment effect among the entire population (ATE), among the treated (ATT), and among the controls (ATC). If the outcome is binary, also the relative risk and odds ratio parameters. P-values and 95% confidence intervals are also calculated (on the log scale for RR and OR).
calcParameters(Y, A, I.Z, Delta, g1W, g0W, Q, mu1, mu0, id, family,
obsWeights, alpha.sig=0.05, ICflag=TRUE)
Population mean outcome estimate, variance, p-value, 95% confidence interval (missingness only, no treatment assignment), or NULL
additive treatment effect estimate, variance, p-value, 95% confidence interval, or NULL
relative risk estimate, p-value, 95% confidence interval, log(RR), variance(log(RR)), or NULL
odds ratio estimate, p-value, 95% confidence interval, log(OR), variance(log(OR)), or NULL
continuous or binary outcome variable
binary treatment indicator, 1 - treatment, 0 - control
Indicator Z=z, needed for CDE estimation
indicator of missing outcome. 1 - observed, 0 - missing
censoring mechanism estimates, \(P(A=1|W) \times P(Delta=1|A,W)\)
censoring mechanism estimates, \(P(A=0|W) \times P(Delta=1|A,W)\)
a 3-column matrix (Q(A,W), Q(1,W), Q(0,W))
targeted estimate of \(E(Y|A=1,W)\)
targeted estimate of \(E(Y|A=0,W)\)
subject identifier
family specification for regressions, generally ‘gaussian’ for continuous outcomes, ‘binomial’ for binary outcomes
sampling weights
significance level for constructing CIs. Default = 0.05
set to FALSE to skip evaluating IC-based variance
Susan Gruber
tmle,
estimateQ,
estimateG,
tmleMSM,
calcSigma