DTAplots (version 1.0.2.5)

SROC_rjags: A function to create a summary plot in Receiver Operating Characteristic (ROC) space

Description

This function creates : 1) a plot of the mean sensitivity and mean specificity together with prediction and credible region resulting from a bivariate meta-analysis model, or 2) a summary ROC curve resulting from a hierarchical summary receiver operating characteristic (HSROC) meta-analysis model. It uses posterior samples of ? from these meta-analysis models obtained with rjags package from a meta-analysis bivariate model for diagnostic test accuracy. It allows for the possibility that the reference test is not perfect.

Usage

SROC_rjags(X, n, model = "Bivariate",
dataset = NULL, ref_std=NULL, title = "Summary plot",
xlab = "Specificity", ylab = "Sensitivity",
Se.range=c(0,1), Sp.range=c(0,1),
SROC_curve = FALSE, lwd.curve=3, col.curve="black", lty.curve="solid",
cred_region = TRUE, predict_region = TRUE,
region_level = 0.95, lty.cred.region = "solid",
lty.predict.region = "dotted", trunc_low = 0.025, trunc_up = 0.025,
col.predict.region = "black", col.cred.region = "red",
lwd.cred.region = 2.5, lwd.predict.region = 2.5,
study_col1 = "blue", study_col2 = rgb(0, 0, 1, 0.15),
study_symbol = "circles", summary.point=TRUE,
pch.summary.point = 19, cex.summary.point = 3,
col.summary.point = 1)

Arguments

X

An mcmc.list object. See details for further explanations.

n

Number of studies in the meta-analysis.

model

specify the model used to create X. Default is "Bivariate". Other option is "HSROC". See details for further explanations.

dataset

a matrix with the number of rows equal to the number of studies and 4 columns. Each row consists of the entries of the 2x2 table of the index test (i.e. test under evaluation) vs. the reference test reported in each study. The ordering of the columns is ++, +-, -+, --, where the first entry refers to the result of the test under evaluation and the second entry refers to the result of the reference test. If set to NULL (default), individual posterior estimates of test under evaluation characteristics, i.e. points with coordinate (sensitivity, (1-specificity)), will not be drawn on the SROC plot.

ref_std

if "dataset" is not set to NULL, then "ref_std" must be set to TRUE if the meta-analysis model assumes a perfect reference standard test or to FALSE if the meta-analysis model assumes an imperfect reference standard test.

title

plot title.

xlab

x axis label.

ylab

y axis label.

Se.range

the y-axis limits of the plot.

Sp.range

the x-axis limits of the plot.

SROC_curve

logical. If TRUE, a SROC curve is drawn.

lwd.curve

The line width of the SROC curve, a positive number, defaulting to 3. See "lwd" in "par".

col.curve

A specification for the default color of the SROC curve. Defaults to "black". See "col" in "par".

lty.curve

The SROC curve line type. Default to "solid". See "lty" in "par".

cred_region

logical. If TRUE (default), a credible region curve is drawn on the plot.

predict_region

logical. If TRUE (default), a prediction region curve is drawn on the plot.

region_level

The credible (prediction) level required for the credible (prediction) region. Default value ie 0.95.

lty.cred.region

The credible region line type. Default to "solid". See "lty" in "par".

lty.predict.region

The prediction region line type. Default to "dotted". See "lty" in "par".

trunc_low

Quantile to determine lower limit of threshold THETA used for creating the SROC curve. Default to 0.025. See details for further explanations.

trunc_up

Quantile to determine upper limit of the threshold THETA used for creating the SROC curve. Default to 0.025. See details for further explanations.

col.predict.region

specification for the default plotting color of the prediction region curve. Defaults to "black".

col.cred.region

specification for the default plotting color of the credible region curve. Defaults to "red".

lwd.cred.region

The line width of the credible region curve, a positive number, defaulting to 2.5.

lwd.predict.region

line width of the prediction region, a positive number, defaulting to 2.5.

study_col1

specification for the default plotting color of the individual study point perimeter. Defaults to "blue".

study_col2

specification for the default plotting color of the individual study point shading. Defaults to rgb(0, 0, 1, 0.15).

study_symbol

specification of the type of symbol to represent the individual study point. Either "circles" (default) or "squares".

summary.point

logical. If TRUE (default), a summary point estimate of coordinate (mean specificity, mean sensitivity) is drawn.

pch.summary.point

An integer specifying a symbol to represent the joint mean sensitivity and mean specificity summary point estimate. See "pch" in "par".

cex.summary.point

The magnification to be used for the symbol to represent the joint mean sensitivity and mean specificity summary point estimate.

col.summary.point

A specification for the default color.

Value

No return value

Details

If model is set to Bivariate, the function expects that the parameters of the bivariate model have been spelled exactly as follows in the jags model :

“mu[1]” for the mean logit-transformed sensitivity. “mu[2]” for the mean logit-transformed specificity. “tau[1]” for the between-study standard deviation in the logit-transformed sensitivity. “tau[2]” for the between-study standard deviation in the logit-transformed specificity. “rho” for the correlation between the logit-transformed sensitivity and the mean logit-transformed specificity. “se” for the logit-transformed sensitivity in individual studies. “sp” for the logit-transformed specificity in individual studies. “Summary_Se” for the sensitivity summary estimate. “Summary_Sp” for the specificity summary estimate.

If model is set to HSROC, the function expects that the parameters of the HSROC model have been spelled exactly as follows in the jags model :

“THETA” for the overall mean cut-off value for defining a positive test (threshold parameter). “LAMBDA” for the overall difference in means (accuracy parameter). “beta” for the logarithm of the ratio of the standard deviation of test results among patients with and without the disease. “tau[1]” for the standard deviation of the accuracy parameter. “tau[2]” for the standard deviation of the threshold parameter. “se” for the logit-transformed sensitivity in individual studies. “sp” for the logit-transformed specificity in individual studies.

The default values for arguments of the function assume the posterior densities provided are for parameters of a Bivariate model. Therefore, it will create a plot with a summary point, credible and prediction region curves.

If the argument SROC_curve is set to true, the function will plot Sensitivity vs Specificity at the posterior mean values of LAMBDA and beta across values of the posterior sample of THETA. The trunc_low and trunc_up arguments truncate the posterior sample of THETA by (trunc_low)*100 to (trunc_up)*100 to ensure the SROC curve remains within the observed scope of the study points.

References

Rutter, C. M., and Gatsonis, C. A. (2001) A hierarchical regression approach to meta-analysis of diagnostic accuracy evaluations. Statistics in Medicine, 20(19):2865-2884.

Reitsma, J.B., Glas, A.S., Rutjes, A.W.S., Scholten, R.J.P.M., Bossuyt, P.M. and Zwinderman, A.H. (2005). Bivariate analysis of sensitivity and specificity produces informative summary measures in diagnostic reviews. Journal of Clinical Epideiology 58, 982-90

Harbord, R.M., Deeks, J.J, Egger, M., Whiting, P., and Sterne, J.A. (2007) A unification of models for meta-analysis of diagnostic accuracy studies. Biostatistics 8, 239-251.

Dendukuri, N., Schiller, I., Joseph, L., and Pai, M. (2012) Bayesian meta-analysis of the accuracy of a test for tuberculosis pleuritis in the absence of a gold-standard reference. Biometrics. doi:10.1111/j. 1541-0420.2012.01773.x

Examples

# NOT RUN {
##############################################################################################
# The example requires the rjags output from a bivariate meta-analysis model.
# We use the posterior_samples_Bivariate dataset which contains the results  of a
# bivariate meta-analysis of Anti_CCP dataset.

#The number of studies
data(Anti_CCP)
n=dim(Anti_CCP)[1]

#We load the samples of the posterior distribution
data(posterior_samples_Bivariate)

# Summary plot
SROC_rjags(X=posterior_samples_Bivariate, model="Bivariate",n=n, study_col1="blue",
           study_col2=rgb(0, 0, 1, 0.15), dataset=Anti_CCP[,2:5], ref_std=TRUE)


##########################################################################################
# The example requires the rjags output from a HSROC meta-analysis model.
# We use the posterior_samples_HSROC dataset which contains the results of a
# HSROC meta-analysis of the Anti_CCP dataset.

#The number of studies
data(Anti_CCP)
n=dim(Anti_CCP)[1]

#We load the samples of the posterior distribution
data(posterior_samples_HSROC)

# Summary plot
SROC_rjags(X=posterior_samples_HSROC, model="HSROC",n=n, study_col1="blue",
           study_col2=rgb(0, 0, 1, 0.15), dataset=Anti_CCP[,2:5], ref_std=TRUE,
           SROC_curve = TRUE, cred_region = FALSE, predict_region = FALSE)

##########################################################################################
# The example requires the rjags output from a Bayesian latent class meta-analysis model.
# We use the posterior_samples_LC dataset which contains the results of a
# latent class meta-analysis of the Xpert dataset.

#The number of studies
data(Xpert)
n=dim(Xpert)[1]

#We load the samples of the posterior distribution
data(posterior_samples_LC)

# Summary plot
SROC_rjags(X=posterior_samples_LC, model="Bivariate",n=n, study_col1="blue",
           study_col2=rgb(0, 0, 1, 0.15), dataset=Xpert[,2:5], ref_std=FALSE)
# }