lavTestWald problem

[Jeremy Miles]

I'm running into a problem with lavTestWald.

It says:

Error in UseMethod("vcov") : 

  no applicable method for 'vcov' applied to an object of class "lavaan"

Even when I run the example in help.

Here's my full session transcript.

> library(lavaan)

Loading required package: stats4

Loading required package: mnormt

Loading required package: pbivnorm

Loading required package: quadprog

Loading required package: numDeriv

This is lavaan 0.5-23.1097

lavaan is BETA software! Please report any bugs.

> data(HolzingerSwineford1939)

> HS.model <- '

+     visual  =~ x1 + b1*x2 + x3

+     textual =~ x4 + b2*x5 + x6

+     speed   =~ x7 + b3*x8 + x9

+ '

> 

> fit <- cfa(HS.model, data=HolzingerSwineford1939)

> 

> # test 1: test about a single parameter

> # this is the 'chi-square' version of the 

> # z-test from the summary() output

> lavTestWald(fit, constraints = "b1 == 0")

Error in UseMethod("vcov") : 

  no applicable method for 'vcov' applied to an object of class "lavaan"

Weirdly, if I patch in the lavTestWald function from github, it works:

> lavTestWald(fit, constraints = "b1 == 1")

$stat

[1] 20.0704

$df

[1] 1

$p.value

[1] 7.464313e-06

$se

[1] "standard"

> lavaan::lavTestWald(fit, constraints = "b1 == 1")

Error in UseMethod("vcov") : 

  no applicable method for 'vcov' applied to an object of class "lavaan"

Even more weirdly, if I copy the contents of the function from typing:

lavTestWald,

call it ltw2 and run that, it also works.

ltw2 <- function (object, constraints = NULL, verbose = FALSE) 

{

    if (object@optim$npar > 0L && !object@optim$converged) 

        stop("lavaan ERROR: model did not converge")

    if (is.null(constraints) || nchar(constraints) == 0L) {

        stop("lavaan ERROR: constraints are empty")

    }

    PT <- as.data.frame(object@ParTable, stringsAsFactors = FALSE)

    eq.idx <- which(PT$op == "==")

    if (length(eq.idx) > 0L) {

        PT <- PT[-eq.idx, ]

    }

    partable <- as.list(PT)

    FLAT <- lavParseModelString(constraints)

    CON <- attr(FLAT, "constraints")

    LIST <- list()

    if (length(CON) > 0L) {

        lhs = unlist(lapply(CON, "[[", "lhs"))

        op = unlist(lapply(CON, "[[", "op"))

        rhs = unlist(lapply(CON, "[[", "rhs"))

        LIST$lhs <- c(LIST$lhs, lhs)

        LIST$op <- c(LIST$op, op)

        LIST$rhs <- c(LIST$rhs, rhs)

    }

    else {

        stop("lavaan ERROR: no equality constraints found in constraints argument")

    }

    theta <- object@optim$x

    ceq.function <- lav_partable_constraints_ceq(partable = partable, 

        con = LIST, debug = FALSE)

    JAC <- try(lav_func_jacobian_complex(func = ceq.function, 

        x = theta), silent = TRUE)

    if (inherits(JAC, "try-error")) {

        JAC <- lav_func_jacobian_simple(func = ceq.function, 

            x = theta)

    }

    if (verbose) {

        cat("Restriction matrix (jacobian):\n")

        print(JAC)

        cat("\n")

    }

    theta.r <- ceq.function(theta)

    if (verbose) {

        cat("Restricted theta values:\n")

        print(theta.r)

        cat("\n")

    }

    VCOV <- vcov(object, labels = FALSE)

    info.r <- JAC %*% VCOV %*% t(JAC)

    Wald <- as.numeric(t(theta.r) %*% solve(info.r) %*% theta.r)

    Wald.df <- nrow(JAC)

    Wald.pvalue <- 1 - pchisq(Wald, df = Wald.df)

    list(stat = Wald, df = Wald.df, p.value = Wald.pvalue, se = object@Options$se)

}

ltw2(fit, constraints = "b1 == 1" )

> ltw2(fit, constraints = "b1 == 1" )

$stat

[1] 20.0704

$df

[1] 1

$p.value

[1] 7.464313e-06

$se

[1] "standard"

>

--

--

My employer has nothing to do with this email. 

[Yves Rosseel]

On 03/02/2017 09:02 PM, Jeremy Miles wrote:
> I'm running into a problem with lavTestWald.
>
> It says:
>
> Error in UseMethod("vcov") :
> no applicable method for 'vcov' applied to an object of class "lavaan"

It would seem that the dispatching of 'vcov' failed for some reason. I
have no explanation for this. Maybe, a variable 'vcov' was created? Or
something else replaced the generic of vcov? It must be something local
as I can not reproduce this on any machine.

In any case, I just changed lavTestWald() so that vcov() is replaced by
a lavaan-specific version, avoiding the dispatch. In dev lavaan 0.6-1.1110

Yves.