ctCheckFit: Visual model fit diagnostics for ctsem fit objects.
In ctsem: Continuous Time Structural Equation Modelling

ctCheckFit

R Documentation

Visual model fit diagnostics for ctsem fit objects.

Description

Visual model fit diagnostics for ctsem fit objects.

Usage

ctCheckFit(
  fit,
  data = TRUE,
  postpred = TRUE,
  priorpred = FALSE,
  statepred = FALSE,
  residuals = FALSE,
  by = fit$ctstanmodelbase$timeName,
  TIpredNames = fit$ctstanmodelbase$TIpredNames,
  nsamples = 30,
  covplot = FALSE,
  corr = TRUE,
  combinevars = NA,
  fastcov = FALSE,
  lagcovplot = FALSE,
  aggfunc = mean,
  aggregate = FALSE,
  groupbysplit = FALSE,
  byNA = TRUE,
  lag = 0,
  smooth = TRUE,
  k = 4,
  breaks = 4,
  entropy = FALSE,
  reg = FALSE,
  verbose = 0,
  indlines = 30
)

Arguments

`fit`	ctStanFit object.
`data`	Include empirical data in plots?
`postpred`	Include post predictive (conditional on estimated parameters and covariates) distribution data in plots?
`priorpred`	Include prior predictive (conditional on priors) distribution data in plots?
`statepred`	Include one step ahead (conditional on estimated parameters, covariates, and earlier data points) distribution data in plots?
`residuals`	Include one step ahead error (conditional on estimated parameters, covariates, and earlier data points) in plots?
`by`	Variable name to split or plot by. 'time', 'LogLik', and 'WhichObs' are also possibilities.
`TIpredNames`	Since time independent predictors do not change with time, by default observations after the first are ignored. For observing attrition it can be helpful to set this to NULL, or when the combinevars argument is used, specifying different names may be useful.
`nsamples`	Number of samples (when applicable) to include in plots.
`covplot`	Splits variables in the model by the 'by' argument, according to the number of breaks (breaks argument), and shows the covariance (or correlation) for the different data sources selected, as well as the differences between each pair.
`corr`	Turns the covplot into a correlation plot. Usually easier to make sense of visually.
`combinevars`	Can be a list of (possibly new) variable names, where each named element of the list contains a character vector of one or more variable names in the fit object, to combine into the one variable. By default, the mean is used, but see the aggfunc argument. The combinevars argument can also be used to ensure that only certain variables are plotted.
`fastcov`	Uses base R cov function for computing covariances. Not recommended with missing data.
`lagcovplot`	Logical. Output lagged covariance type plots?
`aggfunc`	Function to use for aggregation, if needed.
`aggregate`	If TRUE, duplicate observation types are aggregated over using aggfunc. For example, if by = 'time' and there are 8 time points per subject, but breaks = 2, there will be 4 duplicate observation types per 'row' that will be collapsed. In most cases it is helpful to not collapse.
`groupbysplit`	Logical. Affects variable ordering in covariance plots. Defaults to FALSE, grouping by variable, and within variable by split.
`byNA`	Logical. Create an extra break for when the split variable is missing?
`lag`	Integer vector. lag = 1 creates additional variables for plotting, prefixed by 'lag1_', containing the prior row of observations for that subject.
`smooth`	For bivariate plots, use a smoother for estimation?
`k`	Integer denoting number of knots to use in the smoothing spline.
`breaks`	Integer denoting number of discrete breaks to split variables by (when covariance plotting).
`entropy`	Still in development.
`reg`	Logical. Use regularisation when estimating covariance matrices? Can be necessary / faster for some problems.
`verbose`	Logical. If TRUE, shows optimization output when estimating covariances.
`indlines`	Integer number of individual subject lines to draw per data type.