Skip to contents

Fit pooled, subgroup and person-specific linear models

Source: R/stats_fit_lm.R
fit_lm.Rd

Fits a linear model at every requested scope and returns them in one tidy object, so pooled and person-specific results are directly comparable.

Usage

fit_lm(
  data,
  y,
  x,
  id,
  time = NULL,
  scope = "both",
  subgroup = NULL,
  estimator = "native",
  weights = NULL,
  test_prop = 0.2,
  min_train = 10L,
  min_test = 1L,
  ...
)

Arguments

data

Data frame.

y

Outcome column name.

x

Predictors: names, numeric positions, a:b range, formula, or data frame.

id

Person/unit ID column.

time

Optional ordering column.

scope

"both" (pooled + individual), "pooled", "individual", "subgroup", or "all" (pooled + subgroup + individual).

subgroup

Optional subgroup mapping: an find_subgroups() result, a grouping column in data, or a named vector of labels per person.

estimator

"native" for stats::lm(), or "robust" for an M-estimator (MASS::rlm()) that is not dragged around by outliers.

weights

Optional column name in data holding case weights.

test_prop

Proportion of each person's ordered rows held out.

min_train

Minimum complete training rows per person.

min_test

Minimum complete held-out rows per person.

...

Passed to the underlying fitter.

Value

An idiographic_fit.

Examples

fit <- fit_lm(srl, y = "effort", x = "efficacy:monitoring", id = "name",
              time = "day")
metrics(fit, overall = TRUE)
#>        scope model estimator  subject subgroup    n     rmse      mae      bias
#> 1     pooled    lm    native .overall     .all 1150 20.18702 16.28579 0.7323279
#> 2 individual    lm    native .overall     .all 1150 17.23228 12.86623 0.7977109
#>   r_squared
#> 1 0.4353275
#> 2 0.5885302

# A few wild days should not decide a person's slope.
robust <- fit_lm(srl, y = "effort", x = "efficacy:monitoring", id = "name",
                 time = "day", estimator = "robust")
metrics(robust, overall = TRUE)
#>        scope model estimator  subject subgroup    n     rmse      mae      bias
#> 1     pooled    lm    robust .overall     .all 1150 20.09225 16.09355 0.2580218
#> 2 individual    lm    robust .overall     .all 1150 17.30642 12.67565 0.5440047
#>   r_squared
#> 1 0.4406164
#> 2 0.5849822