Package 'clust.bin.pair'

Statistical test for clustered binary matched pair data

Description

A single interface for several adjustments to the mcnemar test for marginal homogeneity that correct for clustered data.

Usage

clust.bin.pair(ak, bk, ck, dk, method = "yang")

Arguments

ak	vector containing counts per group of Success/Success results.
bk	vector containing counts per group of Success/Failure results.
ck	vector containing counts per group of Failure/Success results.
dk	vector containing counts per group of Failure/Failure results.
method	a character string specifying the method to calculate the statistic. Must be one of "yang" (default), "durkalski", "obuchowski", "eliasziw". A value of "mcnemar" can also be supplied for comparison.

Value

A list with class "htest" containing the following components:

statistic	the value of the test statistic.
p.value	the p-value for the test.
method	the type of test applied.
data.name	a character string giving the names of the data.

References

McNemar, Q. (1947). Note on the sampling error of the difference between correlated proportions or percentages. Psychometrika, 12(2), 153-157.

Eliasziw, M., & Donner, A. (1991). Application of the McNemar test to non-independent matched pair data. Statistics in medicine, 10(12), 1981-1991.

Obuchowski, N. A. (1998). On the comparison of correlated proportions for clustered data. Statistics in medicine, 17(13), 1495-1507.

Durkalski, V. L., Palesch, Y. Y., Lipsitz, S. R., & Rust, P. F. (2003). Analysis of clustered matched-pair data. Statistics in medicine, 22(15), 2417-2428.

Yang, Z., Sun, X., & Hardin, J. W. (2010). A note on the tests for clustered matched-pair binary data. Biometrical journal, 52(5), 638-652.

Examples

with(psychiatry, clust.bin.pair(ah, bh, ch, dh, method="eliasziw"))

tc <- nested.to.contingency(thyroids$x.pet, thyroids$x.spect)
clust.bin.pair(tc$ak, tc$bk, tc$ck, tc$dk, method="obuchowski")

oc <- with(obfuscation, paired.to.contingency(group = list(subject, atom),
                                              t1 = control, t2 = treatment))
clust.bin.pair(oc$ak, oc$bk, oc$ck, oc$dk, method="durkalski")

---

Convert between nested results and the canonical contingency tables

Description

Sum all concordant and discordant pairs from each nested group into a contingency table.

Usage

nested.to.contingency(t1, t2)

Arguments

t1	lists of pre-treatment measures
t2	lists of post-treatment measures

Value

Contingency tables represented in the rows of a matrix

Examples

nested.to.contingency(thyroids$x.pet, thyroids$x.spect)

---

Obfuscated C code misinterpretation data

Description

Data from Gopstein et. al.'s experiment on the misinterpretation of C code. Subjects were asked to hand evaluate pairs of functionally equivalent code. Half of the questions were intentionally obfuscated to elicit confusion.

Usage

data(obfuscation)

Format

A data frame with 57 rows and 4 variables:

subject

the ID of the study participant

atom

the type of obfuscation being evaluated

control

whether the subject answered the un-obfuscated question correctly

treatment

whether the subject answered the obfuscated question correctly

Source

Atoms of Confusion

Examples

data(obfuscation)

oc <- paired.to.contingency(group = obfuscation[,c("subject", "atom")],
                            t1    = obfuscation$control,
                            t2    = obfuscation$treatment)

clust.bin.pair(oc$ak, oc$bk, oc$ck, oc$dk, method="durkalski")

---

Convert between paired results and the canonical contingency tables

Description

Group results by common clustering then tally the concordant and discordant pairs.

Usage

paired.to.contingency(group, t1, t2)

Arguments

group	List of grouping values
t1	pre-treatment measures
t2	post-treatment measures

Value

Contingency tables represented in the rows of a matrix

Examples

paired.to.contingency(list(obfuscation$subject, obfuscation$atom),
                      obfuscation$control, obfuscation$treatment)

---

Psychiatrist and patient disagreement data

Description

Psychiatrists and their patients were surveyed in pairs regarding patient concerns and treatment. Each psychiatrist was asked whether each question item was relevant to their patient and each of their patients were asked the same. The data can be evaluated to answer the question of whether there was patient/doctor agreement in each item. The sample was 29 psychiatrists, each with 1-8 patients, for a total of N = 135 matched pairs.

Usage

data(psychiatry)

Format

A data frame with 29 rows and 7 variables:

psychiatrist

the ID of the psychiatrist

Nh

the number of the psychiatrist's patients participating in the experiment

ah

both participants answered 1

bh

patient answered 1, psychiatrist answered 0

ch

patient answered 0, psychiatrist answered 1

dh

both participants answered 0

Wh

Normalized difference: (bh - ch) / Nh

Source

Donner, A., & Petryshen, P. (1989). The statistical analysis of matched data in psychiatric research. Psychiatry research, 28(1), 41-46.

References

Eliasziw, M., & Donner, A. (1991). Application of the McNemar test to non-independent matched pair data. Statistics in medicine, 10(12), 1981-1991.

Examples

data(psychiatry)

psychiatry$Wh == round((psychiatry$bh - psychiatry$ch) / psychiatry$Nh, 2)

clust.bin.pair(psychiatry$ah, psychiatry$bh, psychiatry$ch, psychiatry$dh, method="eliasziw")

---

PET and SPECT data for diagnosing hyperparathyroidism

Description

Following surgery which confirmed the absence of hyperparathyroidism two diagnostic tests, PET and SPECT, were performed. Their measures of true negatives and false positives are reported. Data reported in Obuchowki 1998.

Usage

data(thyroids)

Format

A data frame with 21 rows and 6 variables:

patient

ID of the patient

n.glands

number of glands tested from the patient

n.pet

number of true negatives from the PET test

x.pet

individual results per gland from the PET test

n.spect

number of true negatives from the SPECT test

x.spect

individual results per gland from the SPECT test

Source

Obuchowski, N. A. (1998). On the comparison of correlated proportions for clustered data. Statistics in medicine, 17(13), 1495-1507.

Examples

data(thyroids)

thyroids$n.glands == sapply(thyroids$x.pet, length)
thyroids$n.glands == sapply(thyroids$x.spect, length)

thyroids$n.pet   == sapply(thyroids$x.pet,   function(x) length(which(x == 1)))
thyroids$n.spect == sapply(thyroids$x.spect, function(x) length(which(x == 1)))

tc <- nested.to.contingency(thyroids$x.pet, thyroids$x.spect)

clust.bin.pair(tc[,'ak'], tc[,'bk'], tc[,'ck'], tc[,'dk'], method="obuchowski")

do.call(clust.bin.pair, data.frame(tc))

---