an rms fit object, or for print the result of Predict. options(datadist="d") must have been specified (where d was created by datadist), or it must have been in effect when the the model was fitted.

One or more variables to vary, or single-valued adjustment values. Specify a variable name without an equal sign to use the default display range, or any range you choose (e.g. seq(0,100,by=2),c(2,3,7,14)). The default list of values for which predictions are made is taken as the list of unique values of the variable if they number fewer than 11. For variables with >10 unique values, np equally spaced values in the range are used for plotting if the range is not specified. Variables not specified are set to the default adjustment value limits[2], i.e. the median for continuous variables and a reference category for non-continuous ones. Later variables define adjustment settings. For categorical variables, specify the class labels in quotes when specifying variable values. If the levels of a categorical variable are numeric, you may omit the quotes. For variables not described using datadist, you must specify explicit ranges and adjustment settings for predictors that were in the model. If no variables are specified in ..., predictions will be made by separately varying all predictors in the model over their default range, holding the other predictors at their adjustment values. This has the same effect as specifying name as a vector containing all the predictors. For rbind, ... represents a series of results from Predict. If you name the results, these names will be taken as the values of the new .set. variable added to the concatenated data frames. See an example below.

an optional transformation of the linear predictor. Specify fun='mean' if the fit is a proportional odds model fit and you ran bootcov with coef.reps=TRUE. This will let the mean function be re-estimated for each bootstrap rep to properly account for all sources of uncertainty in estimating the mean response. fun can be a general function and can compute confidence limits (stored as a list in the limits attribute) of the transformed parameters such as means.

set to FALSE if fun is not a function such as the result of Mean, Quantile, or ExProb that contains an intercepts argument

defaults to providing predictions. Set to "model.frame" to return a data frame of predictor settings used. Set to "x" to return the corresponding design matrix constructed from the predictor settings.

the number of equally-spaced points computed for continuous predictors that vary, i.e., when the specified value is . or NA

confidence level (highest posterior density interval probability for Bayesian models). Default is 0.95. Specify FALSE to suppress.

type of confidence interval. Default is "mean" which applies to all models. For models containing a residual variance (e.g, ols), you can specify conf.type="individual" instead, to obtain limits on the predicted value for an individual subject. Specify conf.type="simultaneous" to obtain simultaneous confidence bands for mean predictions with family-wise coverage of conf.int.

set to FALSE to force the use of the bootstrap covariance matrix estimator even when bootstrap coefficient reps are present

set to 'bca' to compute BCa confidence limits or 'basic' to use the basic bootstrap. The default is to compute percentile intervals

defaults to using the posterior mean of the regression coefficients. Specify 'mode' or 'median' to instead use the other summaries.

Set to TRUE to adjust all non-plotted variables to 0 (or reference cell for categorical variables) and to omit intercept(s) from consideration. Default is FALSE.

Set to TRUE to subtract a constant from X beta before plotting so that the reference value of the x-variable yields y=0. This is done before applying function fun. This is especially useful for Cox models to make the hazard ratio be 1.0 at reference values, and the confidence interval have width zero.

This is only useful in a multiple intercept model such as the ordinal logistic model. There to use to second of three intercepts, for example, specify kint=2. The default is 1 for lrm and the middle intercept corresponding to the median y for orm or blrm. You can specify ycut instead, and the intercept corresponding to Y >= ycut will be used for kint.

for an ordinal model specifies the Y cutoff to use in evaluating departures from proportional odds, when the constrained partial proportional odds model is used. When omitted, ycut is implied by kint. The only time it is absolutely mandatory to specify ycut is when computed an effect (e.g., odds ratio) at a level of the response variable that did not occur in the data. This would only occur when the cppo function given to blrm is a continuous function.

Specify a single time u to cause function survest to be invoked to plot the probability of surviving until time u when the fit is from cph or psm.

Specify loglog=TRUE to plot log[-log(survival)] instead of survival, when time is given.

Controls how “adjust-to” values are plotted. The default is 4 significant digits.

Instead of specifying the variables to vary in the variables (...) list, you can specify one or more variables by specifying a vector of character string variable names in the name argument. Using this mode you cannot specify a list of variable values to use; prediction is done as if you had said e.g. age without the equal sign. Also, interacting factors can only be set to their reference values using this notation.

an alternate way of specifying ..., mainly for use by survplot or gendata. This must be a list with one or more values for each variable listed, with NA values for default ranges.

a list containing one value for one variable, which is mandatory if the model included an offset term. The variable name must match the innermost variable name in the offset term. The single offset is added to all predicted values.

an object created by Predict

If you are concatenating predictor sets using rbind and one or more of the variables were renamed for one or more of the sets, but these new names represent different versions of the same predictors (e.g., using or not using imputation), you can specify a named character vector to rename predictors to a central name. For example, specify rename=c(age.imputed='age', corrected.bp='bp') to rename from old names age.imputed, corrected.bp to age, bp. This happens before concatenation of rows.