Fit the marginal structural model for the sequence of emulated trials
Source:R/generics.R, R/modelling.R
fit_msm.Rd![[Experimental]](figures/lifecycle-experimental.svg)
Arguments
- object
A
trial_sequenceobject- weight_cols
character vector of column names in expanded outcome dataset, ie
outcome_data(object). If multiple columns are specified, the element wise product will be used. SpecifyNULLif no weight columns should be used.- modify_weights
a function to transform the weights (or
NULLfor no transformation). Must take a numeric vector of weights and a vector of positive, finite weights of the same length. See examples for some possible function definitions.Before the outcome marginal structural model can be fit, the outcome model must be specified with
set_outcome_model()and the data must be expanded into the trial sequence withexpand_trials().The model is fit based on the
model_fitterspecified in set_outcome_model using the internalfit_outcome_modelmethod.
Examples
trial_seq_object <- trial_sequence("ITT") |>
set_data(data_censored) |>
set_outcome_model(
adjustment_terms = ~age_s,
followup_time_terms = ~ stats::poly(followup_time, degree = 2)
) |>
set_expansion_options(output = save_to_datatable(), chunk_size = 500) |>
expand_trials() |>
load_expanded_data()
fit_msm(trial_seq_object)
#> Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
#> Trial Sequence Object
#> Estimand: Intention-to-treat
#>
#> Data:
#> - N: 725 observations from 89 patients
#> id period treatment x1 x2 x3 x4 age age_s
#> <int> <int> <num> <num> <num> <int> <num> <num> <num>
#> 1: 1 0 1 1 1.146148362 0 0.7342030 36 0.08333333
#> 2: 1 1 1 1 0.002200337 0 0.7342030 37 0.16666667
#> ---
#> 724: 99 6 1 1 -0.033762356 1 0.5752681 71 3.00000000
#> 725: 99 7 0 0 -1.340496520 1 0.5752681 72 3.08333333
#> outcome censored eligible time_on_regime wt
#> <num> <int> <num> <num> <num>
#> 1: 0 0 1 0 1
#> 2: 0 0 0 1 1
#> ---
#> 724: 0 0 0 1 1
#> 725: 1 0 0 2 1
#>
#> IPW for informative censoring:
#> - No weight model specified
#>
#> Sequence of Trials Data:
#> - Chunk size: 500
#> - Censor at switch: FALSE
#> - First period: 0 | Last period: Inf
#>
#> A TE Datastore Datatable object
#> N: 1558 observations
#> id trial_period followup_time outcome weight treatment age_s
#> <int> <int> <int> <num> <num> <num> <num>
#> 1: 1 0 0 0 1 1 0.08333333
#> 2: 1 0 1 0 1 1 0.08333333
#> ---
#> 1557: 99 0 6 0 1 1 2.50000000
#> 1558: 99 0 7 1 1 0 2.50000000
#> assigned_treatment
#> <num>
#> 1: 1
#> 2: 1
#> ---
#> 1557: 1
#> 1558: 1
#>
#> Outcome model:
#> - Formula: outcome ~ assigned_treatment + age_s + stats::poly(followup_time, degree = 2) + trial_period + I(trial_period^2)
#> - Treatment variable: assigned_treatment
#> - Adjustment variables: age_s
#> - Model fitter type: te_stats_glm_logit
#>
#> Model Summary:
#>
#> term estimate std.error statistic p.value
#> (Intercept) -5.46 0.52 -10.60 3.1e-26
#> assigned_treatment 1.34 0.53 2.50 1.2e-02
#> age_s 0.48 0.34 1.42 1.5e-01
#> stats::poly(followup_time, degree = 2)1 -2.23 14.99 -0.15 8.8e-01
#> stats::poly(followup_time, degree = 2)2 -20.02 14.45 -1.39 1.7e-01
#> trial_period 7.05 0.97 7.23 4.8e-13
#> I(trial_period^2) -7.51 0.54 -13.96 2.8e-44
#> conf.low conf.high
#> -6.47 -4.4
#> 0.29 2.4
#> -0.18 1.2
#> -31.60 27.1
#> -48.34 8.3
#> 5.14 9.0
#> -8.57 -6.5
#>
#> null.deviance df.null logLik AIC BIC deviance df.residual nobs
#> 160 1557 -68.4 151 188 137 1551 1558
#>
#> Outcome data
#> N: 1558 observations from 89 patients in 18 trial periods
#> Periods: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
#> id trial_period followup_time outcome weight treatment age_s
#> <int> <int> <int> <num> <num> <num> <num>
#> 1: 1 0 0 0 1 1 0.08333333
#> 2: 1 0 1 0 1 1 0.08333333
#> ---
#> 1557: 99 0 6 0 1 1 2.50000000
#> 1558: 99 0 7 1 1 0 2.50000000
#> assigned_treatment sample_weight w
#> <num> <num> <num>
#> 1: 1 1 1
#> 2: 1 1 1
#> ---
#> 1557: 1 1 1
#> 1558: 1 1 1
# Using modify_weights functions ----
# returns a function that truncates weights to limits
limit_weight <- function(lower_limit, upper_limit) {
function(w) {
w[w > upper_limit] <- upper_limit
w[w < lower_limit] <- lower_limit
w
}
}
# calculate 1st and 99th percentile limits and truncate
p99_weight <- function(w) {
p99 <- quantile(w, prob = c(0.01, 0.99), type = 1)
limit_weight(p99[1], p99[2])(w)
}
# set all weights to 1
all_ones <- function(w) {
rep(1, length(w))
}
fit_msm(trial_seq_object, modify_weights = limit_weight(0.01, 4))
#> Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
#> Trial Sequence Object
#> Estimand: Intention-to-treat
#>
#> Data:
#> - N: 725 observations from 89 patients
#> id period treatment x1 x2 x3 x4 age age_s
#> <int> <int> <num> <num> <num> <int> <num> <num> <num>
#> 1: 1 0 1 1 1.146148362 0 0.7342030 36 0.08333333
#> 2: 1 1 1 1 0.002200337 0 0.7342030 37 0.16666667
#> ---
#> 724: 99 6 1 1 -0.033762356 1 0.5752681 71 3.00000000
#> 725: 99 7 0 0 -1.340496520 1 0.5752681 72 3.08333333
#> outcome censored eligible time_on_regime wt
#> <num> <int> <num> <num> <num>
#> 1: 0 0 1 0 1
#> 2: 0 0 0 1 1
#> ---
#> 724: 0 0 0 1 1
#> 725: 1 0 0 2 1
#>
#> IPW for informative censoring:
#> - No weight model specified
#>
#> Sequence of Trials Data:
#> - Chunk size: 500
#> - Censor at switch: FALSE
#> - First period: 0 | Last period: Inf
#>
#> A TE Datastore Datatable object
#> N: 1558 observations
#> id trial_period followup_time outcome weight treatment age_s
#> <int> <int> <int> <num> <num> <num> <num>
#> 1: 1 0 0 0 1 1 0.08333333
#> 2: 1 0 1 0 1 1 0.08333333
#> ---
#> 1557: 99 0 6 0 1 1 2.50000000
#> 1558: 99 0 7 1 1 0 2.50000000
#> assigned_treatment
#> <num>
#> 1: 1
#> 2: 1
#> ---
#> 1557: 1
#> 1558: 1
#>
#> Outcome model:
#> - Formula: outcome ~ assigned_treatment + age_s + stats::poly(followup_time, degree = 2) + trial_period + I(trial_period^2)
#> - Treatment variable: assigned_treatment
#> - Adjustment variables: age_s
#> - Model fitter type: te_stats_glm_logit
#>
#> Model Summary:
#>
#> term estimate std.error statistic p.value
#> (Intercept) -5.46 0.52 -10.60 3.1e-26
#> assigned_treatment 1.34 0.53 2.50 1.2e-02
#> age_s 0.48 0.34 1.42 1.5e-01
#> stats::poly(followup_time, degree = 2)1 -2.23 14.99 -0.15 8.8e-01
#> stats::poly(followup_time, degree = 2)2 -20.02 14.45 -1.39 1.7e-01
#> trial_period 7.05 0.97 7.23 4.8e-13
#> I(trial_period^2) -7.51 0.54 -13.96 2.8e-44
#> conf.low conf.high
#> -6.47 -4.4
#> 0.29 2.4
#> -0.18 1.2
#> -31.60 27.1
#> -48.34 8.3
#> 5.14 9.0
#> -8.57 -6.5
#>
#> null.deviance df.null logLik AIC BIC deviance df.residual nobs
#> 160 1557 -68.4 151 188 137 1551 1558
#>
#> Outcome data
#> N: 1558 observations from 89 patients in 18 trial periods
#> Periods: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
#> id trial_period followup_time outcome weight treatment age_s
#> <int> <int> <int> <num> <num> <num> <num>
#> 1: 1 0 0 0 1 1 0.08333333
#> 2: 1 0 1 0 1 1 0.08333333
#> ---
#> 1557: 99 0 6 0 1 1 2.50000000
#> 1558: 99 0 7 1 1 0 2.50000000
#> assigned_treatment sample_weight w
#> <num> <num> <num>
#> 1: 1 1 1
#> 2: 1 1 1
#> ---
#> 1557: 1 1 1
#> 1558: 1 1 1
fit_msm(trial_seq_object, modify_weights = p99_weight)
#> Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
#> Trial Sequence Object
#> Estimand: Intention-to-treat
#>
#> Data:
#> - N: 725 observations from 89 patients
#> id period treatment x1 x2 x3 x4 age age_s
#> <int> <int> <num> <num> <num> <int> <num> <num> <num>
#> 1: 1 0 1 1 1.146148362 0 0.7342030 36 0.08333333
#> 2: 1 1 1 1 0.002200337 0 0.7342030 37 0.16666667
#> ---
#> 724: 99 6 1 1 -0.033762356 1 0.5752681 71 3.00000000
#> 725: 99 7 0 0 -1.340496520 1 0.5752681 72 3.08333333
#> outcome censored eligible time_on_regime wt
#> <num> <int> <num> <num> <num>
#> 1: 0 0 1 0 1
#> 2: 0 0 0 1 1
#> ---
#> 724: 0 0 0 1 1
#> 725: 1 0 0 2 1
#>
#> IPW for informative censoring:
#> - No weight model specified
#>
#> Sequence of Trials Data:
#> - Chunk size: 500
#> - Censor at switch: FALSE
#> - First period: 0 | Last period: Inf
#>
#> A TE Datastore Datatable object
#> N: 1558 observations
#> id trial_period followup_time outcome weight treatment age_s
#> <int> <int> <int> <num> <num> <num> <num>
#> 1: 1 0 0 0 1 1 0.08333333
#> 2: 1 0 1 0 1 1 0.08333333
#> ---
#> 1557: 99 0 6 0 1 1 2.50000000
#> 1558: 99 0 7 1 1 0 2.50000000
#> assigned_treatment
#> <num>
#> 1: 1
#> 2: 1
#> ---
#> 1557: 1
#> 1558: 1
#>
#> Outcome model:
#> - Formula: outcome ~ assigned_treatment + age_s + stats::poly(followup_time, degree = 2) + trial_period + I(trial_period^2)
#> - Treatment variable: assigned_treatment
#> - Adjustment variables: age_s
#> - Model fitter type: te_stats_glm_logit
#>
#> Model Summary:
#>
#> term estimate std.error statistic p.value
#> (Intercept) -5.46 0.52 -10.60 3.1e-26
#> assigned_treatment 1.34 0.53 2.50 1.2e-02
#> age_s 0.48 0.34 1.42 1.5e-01
#> stats::poly(followup_time, degree = 2)1 -2.23 14.99 -0.15 8.8e-01
#> stats::poly(followup_time, degree = 2)2 -20.02 14.45 -1.39 1.7e-01
#> trial_period 7.05 0.97 7.23 4.8e-13
#> I(trial_period^2) -7.51 0.54 -13.96 2.8e-44
#> conf.low conf.high
#> -6.47 -4.4
#> 0.29 2.4
#> -0.18 1.2
#> -31.60 27.1
#> -48.34 8.3
#> 5.14 9.0
#> -8.57 -6.5
#>
#> null.deviance df.null logLik AIC BIC deviance df.residual nobs
#> 160 1557 -68.4 151 188 137 1551 1558
#>
#> Outcome data
#> N: 1558 observations from 89 patients in 18 trial periods
#> Periods: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
#> id trial_period followup_time outcome weight treatment age_s
#> <int> <int> <int> <num> <num> <num> <num>
#> 1: 1 0 0 0 1 1 0.08333333
#> 2: 1 0 1 0 1 1 0.08333333
#> ---
#> 1557: 99 0 6 0 1 1 2.50000000
#> 1558: 99 0 7 1 1 0 2.50000000
#> assigned_treatment sample_weight w
#> <num> <num> <num>
#> 1: 1 1 1
#> 2: 1 1 1
#> ---
#> 1557: 1 1 1
#> 1558: 1 1 1