recogROC: Recognition Receiver Operating Characteristics

In mpt: Multinomial Processing Tree Models

Recognition Receiver Operating Characteristics

Description

In a series of experiments, Broeder and Schuetz (2009) tested the shape of recognition receiver operating characteristics. Participants studied a list of items. In a recognition test, old items intermixed with new ones were presented, and participants had to classify them as old or new. The percentage of old items varied in order to manipulate the response bias.

Wellingerhof (2019) conducted a replication study that was designed to be similar to Experiment 3 in Broeder and Schuetz (2009).

Usage

data(recogROC)

Format

ROCBroeder2009 A data frame consisting of seven variables:

item

factor. Target (old) or distractor (new) item.

resp

a factor giving the response category, old or new.

treeid

an identifier for the single trees of the joint multinomial model.

ptarget1, ptarget3

percentage of target (old) items in Experiment 1 and 3, respectively.

exp1, exp3

the aggregate response frequencies.

ROCreplication A data frame containing 48 observations of five variables:

gender

factor. Participant gender.

age

participant age.

arith

number of mental-arithmetic problems solved.

lexical

number of correct trials in lexical selection task.

y

a matrix of aggregate response frequencies per participant. The column names indicate each of 4 x 5 response categories: hit, miss, false alarm, and correct rejection in the five bias conditions.

Source

Broeder, A., & Schuetz, J. (2009). Recognition ROCs are curvilinear–or are they? On premature arguments against the two-high-threshold model of recognition. Journal of Experimental Psychology: Learning, Memory, and Cognition, 35(3), 587–606. doi: 10.1037/a0015279

Wellingerhof, P. (2019). Signal detection theory vs. 2-high-threshold model in recognition memory: A preregistered replication study. Bachelor thesis. University of Tuebingen, Germany. https://osf.io/hvg4p/

See Also

mpt.

Examples

data(recogROC)

## Two-high-threshold model
s <- mptspec("2HT", .replicates = 5,
                    .restr = list(r1=r, r2=r, r3=r, r4=r, r5=r,
                                  d1=d, d2=d, d3=d, d4=d, d5=d))
m1 <- mpt(s, data = ROCBroeder2009, freqvar = "exp3")
m2 <- mpt(s, data = unname(ROCreplication$y))

## Table 4
rbind(Broeder2009 = c(deviance(m1), coef(m1)),
      Replication = c(deviance(m2), coef(m2)))

## Hit rate and false alarm rate
i.hit <- with(ROCBroeder2009, item == "target" & resp == "old")
i.fa  <- with(ROCBroeder2009, item == "distractor" & resp == "old")

hrfa <- data.frame(
   study = rep(c("Broeder2009", "Replication"), each=5),
   obshr = c((m1$y/m1$n)[i.hit], (m2$y/m2$n)[i.hit]),
   obsfa = c((m1$y/m1$n)[i.fa],  (m2$y/m2$n)[i.fa]),
  predhr = c(m1$pcat[i.hit],     m2$pcat[i.hit]),
  predfa = c(m1$pcat[i.fa],      m2$pcat[i.fa])
)

## ROC, Figure 7
plot(obshr ~ obsfa, hrfa[hrfa$study == "Broeder2009", ],
     xlim=0:1, ylim=0:1, pch=16,
     main="Linear recognition ROCs?",
     ylab="Hit rate", xlab="False alarm rate")
abline(0, 1, lty=2)
lines(predhr ~ predfa, hrfa[hrfa$study == "Broeder2009", ])
points(obshr ~ obsfa, hrfa[hrfa$study == "Replication", ], col = "blue")
lines(predhr ~ predfa, hrfa[hrfa$study == "Replication", ],
      col = "blue")
text(0.45, 0.93, "Replication", col = "blue")
text(0.59, 0.82, "Broeder and Schuetz\n(2009, Exp. 3)")

mpt documentation built on March 23, 2022, 9:06 a.m.