Residuals for a beezdemand_tmb fit

Default returns response residuals (y_on_scale - fitted) on the model's native scale. type = "pearson" divides by the residual SD on the model scale (exp(coef[["logsigma_e"]])). Requesting type = "pearson" with scale = "natural" falls back to type = "response" with a message because a response-scale residual SD is not identified for the exponential/zben variants without a separate variance assumption.

Usage

# S3 method for class 'beezdemand_tmb'
residuals(
  object,
  type = c("response", "pearson"),
  scale = c("model", "natural"),
  level = c("subject", "population"),
  ...
)

Arguments

object

A beezdemand_tmb object.

type

One of "response" (default) or "pearson".

scale

One of "model" (default) or "natural".

level

One of "subject" (default; conditions on the subject random effects) or "population" (random effects set to zero, giving the population-mean values). See predict.beezdemand_tmb().

...

Unused.

Value

Numeric vector of length nobs(object).

See also

fitted.beezdemand_tmb(), augment.beezdemand_tmb().

Examples

# \donttest{
data(apt)
fit <- fit_demand_tmb(apt, equation = "exponential", verbose = 0)
#>   equation='exponential': Dropped 14 zero-consumption observations (146 remaining).
head(residuals(fit))
#> [1] -0.012555331  0.035292336  0.082642866 -0.093642127 -0.047270428
#> [6] -0.001380531
head(residuals(fit, type = "pearson"))
#> [1] -0.087977196  0.247299001  0.579091681 -0.656165250 -0.331231397
#> [6] -0.009673602
# }