mod_eval: Evaluate a model for specified inputs
In ProjectMOSAIC/mosaicModel: Create, Visualize, and Predict with Models

Description Usage Arguments Details Value Examples

Find the model outputs for specified inputs. This is equivalent to the generic predict() function, except it will choose sensible values by default. This simplifies getting a quick look at model values.

mod_eval(model = NULL, data = NULL, append = TRUE,
  interval = c("none", "prediction", "confidence"), nlevels = 2,
  bootstrap = 0, ..., on_training = FALSE)

mod_output(model = NULL, data = NULL, append = TRUE,
  interval = c("none", "prediction", "confidence"), nlevels = 2,
  bootstrap = 0, ..., on_training = FALSE)

`model`	the model to display graphically
`data`	optional set of cases from which to extract levels for explanatory variables
`append`	flag whether to include the inputs to the model along with the calculated model value in the output. Default: `TRUE`.
`interval`	the type of interval to use: "none", "confidence", "prediction". But not all types are available for all model architectures. A standard error and 95% confidence interval are produced in the output
`nlevels`	how many levels to construct for input variables. (default: 3) For quantitative variables, this is a suggestion; an attempt is made to have the levels equally spaced. If you're dissatisfied with the result, use `...` to specify exactly what levels you want for any variable.
`bootstrap`	if > 1, the number of bootstrap trials to run to construct a standard error on the model output for each value of the inputs. This is an alternative to `interval`; you can't use both.
`...`	arguments about or values at which to evaluate the model or the kind of output to be passed along to predict(). Unlike `data =` the variables given in `at =` or `...` will be crossed, so that the evaluation will occur at all combinations of the various levels.
`on_training`	flag whether to use the training data for evaluation. Only needed when there are random terms, e.g. from `rand()`, `shuffle()`, .... See details.

There are four distinct ways to specify the values at which the model is to be evaluated. (1) Look for some "typical values" in the data to create a handful of inputs. This is useful for getting a quick look at what the output of the model looks like. This is the default behavior. (2) Using data = to a dataframe containing the explanatory variables will evaluate the model at all of the cases contained in that dataframe. (3) Setting input variables explicitly by using arguments of the form var_name = values, e.g. sex = "F". If not all input variables are specified in this way, the ones that are not will have values set per (1). All combinations of the various variables will be created. See the nlevels argument. (4) Evaluating the model on the training data. There are two ways to do this. The first is to set the data argument to the same data frame used to train the model. The second is to use the on_training = TRUE argument. These are equivalent unless there is some random component among the explanatory terms, as with mosaic::rand(), mosaic::shuffle() and so on.

A dataframe containing both the explanatory variable inputs and the resulting model output (in the model_value field). This differs from the output of stats::predict(), which for many model classes/architectures may be a vector or matrix.

For mod_eval(), a data frame containing both the inputs to the model and the corresponding outputs. For mod_output() a vector containing only the model outputs.

## Not run: 
mod1 <- lm(wage ~ age * sex + sector, data = mosaicData::CPS85)
mod_eval(mod1)
mod2 <- glm(married == "Married" ~ age + sex * sector,
            data = mosaicData::CPS85, family = "binomial")
mod_eval(mod2, nlevels = 2)
mod_eval(mod2, nlevels = 2, sex = "F")

## End(Not run)