designing custom model objects in R - r

Question

I'm coded up an estimator in R and tried to follow R syntax. It goes something like this:
model <- myEstimator(y ~ x1 + x2, data = df)
model has the usual stuff: coefficients, standard errors, p-values, etc.
Now I want model to play nicely with the R ecosystem for summarizing models, like summary() or sjPlot::plot_model() or stargazer::stargazer(). The way you might do summary(lm_model) where lm_model is an lm object.
How do I achieve this? Is there a standard protocol? Define a custom S3 class? Or just coerce model to an existing class like lm?

Answer 1

Create an S3 class and implement summary etc methods.
myEstimator <- function(formula, data) {
result <- list(
coefficients = 1:3,
residuals = 1:3
)
class(result) <- "myEstimator"
result
}
model <- myEstimator(y ~ x1 + x2, data = df)
Functions like summary will just call summary.default.
summary(model)
#> Length Class Mode
#> coefficients 3 -none- numeric
#> residuals 3 -none- numeric
If you wish to have your own summary function, implement summary.myEstimator.
summary.myEstimator <- function(object, ...) {
value <- paste0(
"Coefficients: ", paste0(object$coefficients, collapse = ", "),
"; Residuals: ", paste0(object$residuals, collapse = ", ")
)
value
}
summary(model)
#> [1] "Coefficients: 1, 2, 3; Residuals: 1, 2, 3"
If your estimator is very similar to lm (your model is-a lm), then just add your class to the lm class.
myEstimatorLm <- function(formula, data) {
result <- lm(formula, data)
# Some customisation
result$coefficients <- pmax(result$coefficients, 1)
class(result) <- c("myEstimatorLm", class(result))
result
}
model_lm <- myEstimatorLm(Petal.Length ~ Sepal.Length + Sepal.Width, data = iris)
class(model_lm)
#> [1] "myEstimator" "lm"
Now, summary.lm will be used.
summary(model_lm)
#> Call:
#> lm(formula = formula, data = data)
#>
#> Residuals:
#> Min 1Q Median 3Q Max
#> -1.25582 -0.46922 -0.05741 0.45530 1.75599
#>
#> Coefficients:
#> Estimate Std. Error t value Pr(>|t|)
#> (Intercept) 1.00000 0.56344 1.775 0.078 .
#> Sepal.Length 1.77559 0.06441 27.569 < 2e-16 ***
#> Sepal.Width 1.00000 0.12236 8.173 1.28e-13 ***
#> ---
#> Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
#>
#> Residual standard error: 0.6465 on 147 degrees of freedom
#> Multiple R-squared: 0.8677, Adjusted R-squared: 0.8659
#> F-statistic: 482 on 2 and 147 DF, p-value: < 2.2e-16
You can still implement summary.myEstimatorLm
summary.myEstimatorLm <- summary.myEstimator
summary(model_lm)
#> [1] "Coefficients: 1, 1.77559254648113, 1; Residuals: ...