FIAR (version 0.6)

ARsem: Auto-regressive SEM

Description

Fit an auto-regressive SEM of the specified order. The function automatically extends the given model and dataset to the model and dataset of the specified order. The function is actually a wrapper around the function sem() from the package lavaan.

Usage

ARsem(model, data, order)

Arguments

model

A vector specifying the model of AR order 0 (AR(0)) . The vector should be written as an n by n matrix where n is the number of regions in the network (see example). For every expected connection ij from region i (column) to region j (row) the vector contains '1' and '0' otherwise.

data

contains all observations (rows) and variables (colums) in the network m0. Only variables that are in the model m0 should be in the dataset.

order

Integer. The order of the AR model.

Value

An object of class 'lavaanModel', for which several methods are available, including a 'summary' method.

Details

An AR model of order q contains the t-0, t-1, t-2,...t-q timeseries and these timeseries are connected based on the model of order 0. The function will transform this model in the correct AR(q) model and the data set in the lagged data set containing all lagged variables. Let us take the very simple example of a dataset with 2 variables X and Y. If there is an arrow from X to Y, the function will create the AR(1) model with an additional arrow from X-1 to Y-1, from X-1 to X and from Y-1 to Y. Variables X-1 and Y-1 are automatically created within the function.

References

Kim, J., Zhu, W., Chang, L., Bentler, P., and Ernst, T. (2007). Unified Structural Equation Modeling Approach for the analysis of Multisubject, Multivariate Functional MRI Data. Human Brain Mapping 85 93.

Examples

# NOT RUN {
# Example dataset with three brainregions x, y, z.
head(semdata)
# Prior model with connections from (column) x to (row) y and from y to z.
model <- c(0,0,0,
           1,0,0,
	   0,1,0)
# Perform classical SEM

 fit0 <- ARsem(model,semdata)
 summary(fit0)

# Calculate AR() order of the data
ARorder(semdata,max=10)

# Compute AR(3) SEM
 fit3 <- ARsem(model,semdata,order=3)
 summary(fit3)
# }