rbmi (version 1.7.1)

ancova: Analysis of Covariance

Description

Performs an analysis of covariance between two or more treatment groups returning the estimated "treatment effect" (i.e. the contrast between treatment groups) and the least square means estimates in each group.

Usage

ancova(
  data,
  vars,
  visits = NULL,
  weights = c("counterfactual", "equal", "proportional_em", "proportional")
)

Value

A named list with one set of entries per visit, each suffixed by the visit name. For each visit the list contains:

• the estimated treatment effect(s). For the default (group_contrasts = NULL) these are trt_<visit> for the alt vs ref comparison and trt_alt2_<visit>, trt_alt3_<visit>, ... for further non-reference groups versus the reference group. When group_contrasts is supplied each contrast is named <name>_<visit> using the list name given for that contrast. Also

• the least square means for each group (lsm_ref_<visit>, lsm_alt_<visit>, lsm_alt2_<visit>, ...).

For the common case of two groups this reduces to trt_<visit>, lsm_ref_<visit>

and lsm_alt_<visit>. Each of these elements is itself a list holding the estimate (est), standard error (se) and degrees of freedom (df).

Arguments

data

A data.frame containing the data to be used in the model.

vars

A vars object as generated by set_vars(). Only the group, visit, outcome and covariates elements are required. See details.

visits

An optional character vector specifying which visits to fit the ANCOVA model at. If NULL, a separate ANCOVA model will be fit to the outcomes for each visit (as determined by unique(data[[vars$visit]])). See details.

weights

Character, either "counterfactual" (default), "equal", "proportional_em" or "proportional". Specifies the weighting strategy to be used when calculating the lsmeans. See the weighting section for more details.

Weighting

Counterfactual

For weights = "counterfactual" (the default) the lsmeans are obtained by taking the average of the predicted values for each patient after assigning all patients to each arm in turn. This approach is equivalent to standardization or g-computation. In comparison to emmeans this approach is equivalent to:

emmeans::emmeans(model, specs = "<treatment>", counterfactual = "<treatment>")

Note that to ensure backwards compatibility with previous versions of rbmi weights = "proportional" is an alias for weights = "counterfactual". To get results consistent with emmeans's weights = "proportional" please use weights = "proportional_em".

Equal

For weights = "equal" the lsmeans are obtained by taking the model fitted value of a hypothetical patient whose covariates are defined as follows:

• Continuous covariates are set to mean(X)

• Dummy categorical variables are set to 1/N where N is the number of levels

• Continuous * continuous interactions are set to mean(X) * mean(Y)

• Continuous * categorical interactions are set to mean(X) * 1/N

• Dummy categorical * categorical interactions are set to 1/N * 1/M

In comparison to emmeans this approach is equivalent to:

emmeans::emmeans(model, specs = "<treatment>", weights = "equal")

Proportional

For weights = "proportional_em" the lsmeans are obtained as per weights = "equal" except instead of weighting each observation equally they are weighted by the proportion in which the given combination of categorical values occurred in the data. In comparison to emmeans this approach is equivalent to:

emmeans::emmeans(model, specs = "<treatment>", weights = "proportional")

Note that this is not to be confused with weights = "proportional" which is an alias for weights = "counterfactual".

Details

The function works as follows:

1. Select the first value from visits.

2. Subset the data to only the observations that occurred on this visit.

3. Fit a linear model as vars$outcome ~ vars$group + vars$covariates.

4. Extract the "treatment effect" & least square means for each treatment group.

5. Repeat points 2-3 for all other values in visits.

If no value for visits is provided then it will be set to unique(data[[vars$visit]]).

In order to meet the formatting standards set by analyse() the results will be collapsed into a single list suffixed by the visit name, e.g.:

list(
   trt_visit_1 = list(est = ...),
   lsm_ref_visit_1 = list(est = ...),
   lsm_alt_visit_1 = list(est = ...),
   trt_visit_2 = list(est = ...),
   lsm_ref_visit_2 = list(est = ...),
   lsm_alt_visit_2 = list(est = ...),
   ...
)

ancova() supports two or more treatment groups. The group levels are referred to via a fixed naming scheme derived from the factor levels of vars$group: ref is the first factor level, alt the second, alt2 the third, alt3 the fourth, and so on. Note that ref does not necessarily coincide with the control arm; it is simply the first factor level.

The least square means for each group are returned as lsm_ref, lsm_alt, lsm_alt2, etc. Treatment effects (model contrasts) are returned as trt for the alt vs ref comparison, trt_alt2 for alt2 vs ref, and so on. For the common case of two groups this reduces to the original trt, lsm_ref and lsm_alt naming, ensuring backwards compatibility.

By default a treatment effect is calculated for every non-reference group versus the reference group, using the trt / trt_alt2 / ... names described above. Alternatively a bespoke set of contrasts can be requested via the group_contrasts argument of set_vars(); see its documentation for details. Such contrasts must be named, and the supplied name is used as the parameter name and carried through to the contrast_label column of the pool() output. Contrasts may be pairwise (a length-2 c(minuend, subtrahend) character vector) or general linear contrasts (a named numeric weight vector over the group levels).

If you want to include interaction terms in your model this can be done by providing them to the covariates argument of set_vars() e.g. set_vars(covariates = c("sex*age")). Interactions involving the treatment group (vars$group) are not supported and will result in an error, because the treatment contrasts would depend on the reference values of the interacting covariates.

Note that the treatment effects (trt, trt_alt2, ...) are the relevant linear combinations of the group main-effect coefficients. The least square means (lsm_*) are computed according to the requested weights.

For treatment contrasts, numeric covariates are set to their observed means within each visit, excluding missing values, before formula transformations are applied (e.g. log(mean(age, na.rm = TRUE)) for log(age)). Factor covariates are set to their first level and logical covariates to FALSE. With treatment-group interactions excluded, these reference values do not affect the treatment contrasts, provided the transformed values are finite.

See Also

analyse()

stats::lm()

set_vars()

Examples

# Simulate a small dataset with a single visit, a treatment group and a
# baseline covariate to adjust for.
set.seed(101)
dat <- data.frame(
    visit = factor("visit_1"),
    group = factor(rep(c("Control", "Intervention"), each = 50)),
    basval = rnorm(100)
)
dat$outcome <- 5 + 2 * (dat$group == "Intervention") + dat$basval + rnorm(100)

vars <- set_vars(
    outcome = "outcome",
    group = "group",
    visit = "visit",
    covariates = "basval"
)

# Estimated treatment effect and least square means for the single visit.
# In a full `rbmi` analysis, `ancova()` is passed to `analyse()` rather than
# called directly; see [analyse()].
ancova(dat, vars)

# Multi-arm ANCOVA with a bespoke, named set of contrasts. With three groups
# the default would compare each active arm against the reference ("Placebo");
# here we additionally request the "High" vs "Low" contrast. Explicit contrasts
# must be named, and the names become the output parameter names.
set.seed(102)
dat3 <- data.frame(
    visit = factor("visit_1"),
    group = factor(
        rep(c("Placebo", "Low", "High"), each = 50),
        levels = c("Placebo", "Low", "High")
    ),
    basval = rnorm(150)
)
dat3$outcome <- 5 +
    2 * (dat3$group == "Low") +
    4 * (dat3$group == "High") +
    dat3$basval +
    rnorm(150)

vars3 <- set_vars(
    outcome = "outcome",
    group = "group",
    visit = "visit",
    covariates = "basval",
    group_contrasts = list(
        low_vs_pbo  = c("Low", "Placebo"),
        high_vs_pbo = c("High", "Placebo"),
        high_vs_low = c("High", "Low")
    )
)

# Output names: `low_vs_pbo`, `high_vs_pbo`, `high_vs_low`, plus
# `lsm_ref` / `lsm_alt` / `lsm_alt2`.
names(ancova(dat3, vars3))