estimate: GGM: Estimation test

In donaldRwilliams/bggmMock: mock

Description

Estimate the conditional (in)dependence with either an analytic solution or efficiently sampling from the posterior distribution. These methods were introduced in \insertCiteWilliams2019;textualbggmMock. The graph is selected with select.estimate and then plotted with plot.select.

Usage

estimate(
  Y,
  formula = NULL,
  type = "continuous",
  mixed_type = NULL,
  analytic = FALSE,
  prior_sd = 0.25,
  iter = 5000,
  progress = TRUE,
  seed = 1,
  ...
)

Arguments

Y	Matrix (or data frame) of dimensions n (observations) by p (variables).
formula	An object of class formula. This allows for including control variables in the model (i.e., ~ gender). See the note for further details.
type	Character string. Which type of data for Y ? The options include continuous, binary, ordinal, or mixed. Note that mixed can be used for data with only ordinal variables. See the note for further details.
mixed_type	Numeric vector. An indicator of length p for which varibles should be treated as ranks. (1 for rank and 0 to assume normality). The default is currently to treat all integer variables as ranks when type = "mixed" and NULL otherwise. See note for further details.
analytic	Logical. Should the analytic solution be computed (default is FALSE)?
prior_sd	Scale of the prior distribution, approximately the standard deviation of a beta distribution (defaults to 0.50).
iter	Number of iterations (posterior samples; defaults to 5000).
progress	Logical. Should a progress bar be included (defaults to TRUE) ?
seed	An integer for the random seed.
...	Currently ignored.

Details

The default is to draw samples from the posterior distribution (analytic = FALSE). The samples are required for computing edge differences (see ggm_compare_estimate), Bayesian R2 introduced in \insertCitegelman_r2_2019;textualbggmMock, etc. If the goal is to *only* determine the non-zero effects, this can be accomplished by setting analytic = TRUE. This is particularly useful when a fast solution is needed (see the examples in ggm_compare_ppc)

Controlling for Variables:

When controlling for variables, it is assumed that Y includes only the nodes in the GGM and the control variables. Internally, only the predictors that are included in formula are removed from Y. This is not behavior of, say, lm, but was adopted to ensure users do not have to write out each variable that should be included in the GGM. An example is provided below.

Mixed Type:

The term "mixed" is somewhat of a misnomer, because the method can be used for data including only continuous or only discrete variables. This is based on the ranked likelihood which requires sampling the ranks for each variable (i.e., the data is not merely transformed to ranks). This is computationally expensive when there are many levels. For example, with continuous data, there are as many ranks as data points!

The option mixed_type allows the user to determine which variable should be treated as ranks and the "emprical" distribution is used otherwise. This is accomplished by specifying an indicator vector of length p. A one indicates to use the ranks, whereas a zero indicates to "ignore" that variable. By default all integer variables are handled as ranks.

Dealing with Errors:

An error is most likely to arise when type = "ordinal". The are two common errors (although still rare):

• The first is due to sampling the thresholds, especially when the data is heavily skewed. This can result in an ill-defined matrix. If this occurs, we recommend to first try decreasing prior_sd (i.e., a more informative prior). If that does not work, then change the data type to type = mixed which then estimates a copula GGM (this method can be used for data containing only ordinal variable). This should work without a problem.

• The second is due to how the ordinal data are categorized. For example, if the error states that the index is out of bounds, this indicates that the first category is a zero. This is not allowed, as the first category must be one. This is addressed by adding one (e.g., Y + 1) to the data matrix.

Value

The returned object of class estimate contains a lot of information that is used for printing and plotting the results. For users of bggmMock, the following are the useful objects:

• pcor_mat Partial correltion matrix (posterior mean).

• post_samp An object containing the posterior samples.

Note

Posterior Uncertainty:

A key feature of bggmMock is that there is a posterior distribution for each partial correlation. This readily allows for visiualizing uncertainty in the estimates. This feature works with all data types and is accomplished by plotting the summary of the estimate object (i.e., plot(summary(fit))). Several examples are provided below.

Interpretation of Conditional (In)dependence Models for Latent Data:

See bggmMock-package for details about interpreting GGMs based on latent data (i.e, all data types besides "continuous")

References

\insertAllCited

Examples

#########################################
### example 1: continuous and ordinal ###
#########################################
# data
Y <- ptsd

# continuous

# fit model
fit <- estimate(Y,
                type = "continuous",
                iter = 10)

donaldRwilliams/bggmMock documentation built on May 14, 2020, 12:33 a.m.