Geom for predictions

This geom serves to visualize prediction objects which usually results from a call to predict.bru(). Predictions objects provide summary statistics (mean, median, sd, ...) for one or more random variables. For single variables (or if requested so by setting bar = TRUE), a boxplot-style geom is constructed to show the statistics. For multivariate predictions the mean of each variable (y-axis) is plotted agains the row number of the varriable in the prediction data frame (x-axis) using geom_line. In addition, a geom_ribbon is used to show the confidence interval.

Note: gg.prediction also understands the format of INLA-style posterior summaries, e.g. fit$summary.fixed for an inla object fit

Requires the ggplot2 package.

Usage

# S3 method for prediction
gg(data, mapping = NULL, ribbon = TRUE, alpha = 0.3, bar = FALSE, ...)

Arguments

data: A prediction object, usually the result of a predict.bru() call.
mapping: a set of aesthetic mappings created by aes. These are passed on to geom_line.
ribbon: If TRUE, plot a ribbon around the line based on the smalles and largest quantiles present in the data, found by matching names starting with q and followed by a numerical value. inla()-style numeric+"quant" names are converted to inlabru style before matching.
alpha: The ribbons numeric alpha (transparency) level in [0,1].
bar: If TRUE plot boxplot-style summary for each variable.
...: Arguments passed on to geom_line.

Value

Concatenation of a geom_line value and optionally a geom_ribbon value.

Examples

# \donttest{
if (bru_safe_inla() &&
    require(sn, quietly = TRUE) &&
    require(ggplot2, quietly = TRUE)) {
  # Generate some data

  input.df <- data.frame(x = cos(1:10))
  input.df <- within(input.df, y <- 5 + 2 * cos(1:10) + rnorm(10, mean = 0, sd = 0.1))

  # Fit a model with fixed effect 'x' and intercept 'Intercept'

  fit <- bru(y ~ x, family = "gaussian", data = input.df)

  # Predict posterior statistics of 'x'

  xpost <- predict(fit, NULL, formula = ~x_latent)

  # The statistics include mean, standard deviation, the 2.5% quantile, the median,
  # the 97.5% quantile, minimum and maximum sample drawn from the posterior as well as
  # the coefficient of variation and the variance.

  xpost

  # For a single variable like 'x' the default plotting method invoked by gg() will
  # show these statisics in a fashion similar to a box plot:
  ggplot() +
    gg(xpost)


  # The predict function can also be used to simultaneously estimate posteriors
  # of multiple variables:

  xipost <- predict(fit,
    newdata = NULL,
    formula = ~ c(
      Intercept = Intercept_latent,
      x = x_latent
    )
  )
  xipost

  # If we still want a plot in the previous style we have to set the bar parameter to TRUE

  p1 <- ggplot() +
    gg(xipost, bar = TRUE)
  p1

  # Note that gg also understands the posterior estimates generated while running INLA

  p2 <- ggplot() +
    gg(fit$summary.fixed, bar = TRUE)
  multiplot(p1, p2)

  # By default, if the prediction has more than one row, gg will plot the column 'mean' against
  # the row index. This is for instance usefuul for predicting and plotting function
  # but not very meaningful given the above example:

  ggplot() +
    gg(xipost)

  # For ease of use we can also type

  plot(xipost)

  # This type of plot will show a ribbon around the mean, which viszualizes the upper and lower
  # quantiles mentioned above (2.5 and 97.5%). Plotting the ribbon can be turned of using the
  # \code{ribbon} parameter

  ggplot() +
    gg(xipost, ribbon = FALSE)

  # Much like the other geomes produced by gg we can adjust the plot using ggplot2 style
  # commands, for instance

  ggplot() +
    gg(xipost) +
    gg(xipost, mapping = aes(y = median), ribbon = FALSE, color = "red")
}


# }

Usage

Arguments

Value

See also

Examples