mnet (version 0.1.4)

VAR_GC: Function to test whether parameters of two VAR(1) models are different

Description

Function to test whether parameters of two VAR(1) models are different

Usage

VAR_GC(data, vars, dayvar, beepvar,
       groups, test = "parametric",
       nP = 1000)

Value

Call

Returns the call of the function

phi_diff

A p x p matrix of differences in phi coefficients (Group 1 - Group 2) in the empirical data. The test-statistics.

phi_pval

A p x p matrix with pvalues corresponding to phi_diff.

int_diff

A p numeric vector of differences in intercepts (Group 1 - Group 2).

int_pvar

A p numeric vector of pvalues corresponding to int_diff

Arguments

data

A n x p data matrix.

vars

An integeger vector indicating the column numbers of the variables that should be modeled in the pair of VAR models.

dayvar

Integener indicating the column number of the variable specifying the day of the measurement point. This is designed for EMA studies. If not applicable, this argument can be left unspecified.

beepvar

Integener indicating the column number of the variable specifying the number of the measurement occasion during a specific day. The full set of measurement occasions has to be consecutive and increasing sequence of integers (e.g., 1, 2,..., 5). This is designed for EMA studies. If not applicable, this argument can be left unspecified.

groups

Integer indiciating the column number of the group variable. The values of the group variable have to be 1 and 2.

test

If test = "parmetric" p-values are computed using the t-distribution, using the Welch-Satterthwaite equation to approximate degrees of freedom (df). This is the default. Alternatively, test = "permutation" uses a permutation to sample from a distribution in which the null hypothesis that no group differences exist is true. The permutation uses nP permuted datasets.

nP

The number of permuted datasets used if test = "permutation".

Author

Jonas Haslbeck <[e-mail]>

Examples

# \donttest{

library(mlVAR) # for simulateVAR() function

# Specify Model
p <- 4
A1 <- diag(p) * 0.8
A2 <- diag(p) * 0.8
A2[2,1] <- 0.7

# Simulate datasets
Nt <- 500
set.seed(13)  # for reproducibility
data1_x <- simulateVAR(A1, means=rep(0, p), Nt = Nt, residuals=.1)
data2_x <- simulateVAR(A2, means=rep(0, p), Nt = Nt, residuals=.1)

# Add beep and day vars
dayvar1 <- dayvar2 <- rep(1:(Nt/5), each=5)
beepvar1 <- beepvar2 <- rep(1:5, Nt/5)

# Add grouping var
groups1 <- rep(1, Nt)
groups2 <- rep(2, Nt)

# Combine
data1 <- data.frame(cbind(dayvar1, beepvar1, groups1, data1_x))
data2 <- data.frame(cbind(dayvar2, beepvar2, groups2, data2_x))
colnames(data1) <- colnames(data2) <- c("dayvar", "beepvar", "groups", paste0("V", 1:4))
data <- rbind(data1, data2)


# Call
out <- VAR_GC(data = data,
              vars = 4:7,
              dayvar = 1,
              beepvar = 2,
              groups = 3)

round(out$phi_pval, 2)
round(out$phi_pval[2,1], 2) # worked!

# }