clm2twoAC: Extract 2-AC coefficient table from a cumulative link model

Description

The Thurstonian model for the 2-AC protocol can be formulated as a cumulative link model (see the references). This function extracts the 2-AC model parameter estimates, standard errors, z-value and p-values from a cumulative link (mixed) model fitted with clm or clmm from package ordinal.

Usage

clm2twoAC(object, ...)

Arguments

object

a clm or clmm object

…

not currently used.

Value

A data.frame with the coefficient table. The two first rows contain the estimates of tau and d.prime while the remaining rows contain optional regression variables for d.prime.

References

Christensen R.H.B., Lee H-S and Brockhoff P.B. (2011). Estimation of the Thurstonian model for the 2-AC protocol. Submitted to Food Quality and Preference.

Examples

Run this code

# NOT RUN {
## Example of a simple 2-AC model. First the conventional way:
twoAC(c(2, 2, 6))
# }
# NOT RUN {
## The using a cumulative link model (clm from package ordinal):
if(require(ordinal)) {
    response <- gl(3,1)
    fit.clm <- clm(response ~ 1, weights = c(2, 2, 6), link = "probit")
    clm2twoAC(fit.clm)
    ## Alternatively we could get estimates and standard errors "by hand":
    tab <- coef(summary(fit.clm))
    theta <- tab[,1]
    (tau <- (theta[2] - theta[1])/sqrt(2))
    (d.prime <- (-theta[2] - theta[1])/sqrt(2))
    VCOV <- vcov(fit.clm)
    (se.tau <- sqrt((VCOV[1,1] + VCOV[2,2] - 2*VCOV[2,1])/2))
    (se.d.prime <- sqrt((VCOV[1,1] + VCOV[2,2] + 2*VCOV[2,1])/2))

    ## Extended example with a regression model for d.prime
    ## (see the referenced paper for details):
    n.women <- c(2, 2, 6)*10
    n.men <- c(1, 2, 7)*10
    wt <- c(n.women, n.men)
    response <- gl(3,1, length = 6)
    gender <- gl(2, 3, labels = c("women", "men"))
    fm2 <- clm(response ~ gender, weights = wt, link = "probit")
    clm2twoAC(fm2)
}
# }
# NOT RUN {
# }