Skip to contents

Plot Loss Surface for Demand Model Parameters

Source: R/plot-viz.R
plot_loss_surface.Rd

Visualizes the sum-of-squared-residuals (SSR) surface over a grid of Q0 and alpha values, holding other parameters (k, variance components) fixed at their MLE. The SSR is computed on aggregated mean log-consumption by price for the Part II (continuous) component of hurdle models.

Usage

plot_loss_surface(object, ...)

# S3 method for class 'beezdemand_hurdle'
plot_loss_surface(
  object,
  resolution = 80,
  q0_range = NULL,
  alpha_range = NULL,
  fill_palette = "D",
  show_mle = TRUE,
  show_contours = FALSE,
  style = c("modern", "apa"),
  ...
)

# S3 method for class 'beezdemand_tmb'
plot_loss_surface(
  object,
  resolution = 80,
  q0_range = NULL,
  alpha_range = NULL,
  fill_palette = "D",
  show_mle = TRUE,
  show_contours = FALSE,
  style = c("modern", "apa"),
  ...
)

# S3 method for class 'beezdemand_nlme'
plot_loss_surface(
  object,
  resolution = 80,
  q0_range = NULL,
  alpha_range = NULL,
  fill_palette = "D",
  show_mle = TRUE,
  show_contours = FALSE,
  style = c("modern", "apa"),
  type = c("ssr", "marginal"),
  ...
)

Arguments

object

A fitted model object.

...

Additional arguments passed to methods.

resolution

Integer; grid resolution per axis (default 80).

q0_range

Numeric vector of length 2; Q0 range (natural scale). Default: MLE +/- 3 orders of magnitude.

alpha_range

Numeric vector of length 2; alpha range (natural scale). Default: MLE +/- 3 orders of magnitude.

fill_palette

Character; viridis palette option (default "D").

show_mle

Logical; overlay MLE point (default TRUE).

show_contours

Logical; add contour lines (default FALSE).

style

Character; plot style, "modern" or "apa".

type

Character; loss surface to plot for NLME models. "ssr" (default) uses sum-of-squared-residuals on price-aggregated means; "marginal" uses a linearized marginal negative log-likelihood.

Value

A ggplot2 object.

Details

Important: This function computes SSR on price-aggregated means, not a true profile likelihood. The resulting surface shows how well different (Q0, alpha) pairs explain the average demand pattern, but does not account for individual variation. For models with large random effects, the surface may appear sharper than the full-data objective.

Supported model classes: beezdemand_hurdle, beezdemand_tmb, and beezdemand_nlme. Models with factor covariates on Q0 or alpha are not supported; use get_demand_param_emms() instead.

See also

plot_loss_profile() for 1D profile slices

Examples

# \donttest{
data(apt)
fit <- fit_demand_hurdle(apt, y_var = "y", x_var = "x", id_var = "id")
#> Sample size may be too small for reliable estimation.
#>   Subjects: 10, Parameters: 12, Recommended minimum: 60 subjects.
#>   Consider using more subjects or the simpler 2-RE model.
#> Fitting HurdleDemand3RE model...
#>   Part II: zhao_exponential
#>   Subjects: 10, Observations: 160
#>   Fixed parameters: 12, Random effects per subject: 3
#>   Optimizing...
#>   Converged in 81 iterations
#>   Computing standard errors...
#> Done. Log-likelihood: 32.81
plot_loss_surface(fit)

plot_loss_surface(fit, resolution = 50, show_contours = TRUE)
#> Warning: The following aesthetics were dropped during statistical transformation: fill.
#> ℹ This can happen when ggplot fails to infer the correct grouping structure in
#>   the data.
#> ℹ Did you forget to specify a `group` aesthetic or to convert a numerical
#>   variable into a factor?

# }