Package 'mrbayes'

mrbayes: Bayesian implementation of the IVW and MR-Egger models for two-sample Mendelian randomization analyses

Description

Bayesian implementation of the IVW and MR-Egger models and their radial and multivariate versions for two-sample Mendelian randomization analyses.

Author(s)

Maintainer: Tom Palmer [email protected] (ORCID)

Authors:

• Okezie Uche-Ikonne [email protected] (ORCID)

• Frank Dondelinger [email protected] (ORCID)

References

Stan Development Team (2019). RStan: the R interface to Stan. R package version 2.19.2. https://mc-stan.org

See Also

Useful links:

• https://github.com/okezie94/mrbayes

• https://okezie94.github.io/mrbayes/

• https://mrcieu.r-universe.dev/mrbayes

• Report bugs at https://github.com/okezie94/mrbayes/issues

---

Dataset from Richmond et. al 2017 investigating the association of BMI on insulin resistance

Description

A two-sample summary level dataset, Richmond et al. (2017) doi:10.1101/155739, containing 14 single nucleiodtide polymorphisms (SNPs) which have genotype-phenotype associations (BMI) and genotype-outcome associations (insulin) with their respective standard errors.

Usage

bmi_insulin

Format

A data frame with 14 rows and 44 columns:

rsid

SNP RSID number

beta.exposure

The genotype-BMI associations

beta.outcome

The genotype-outcome associations

se.exposure

The standard errors of the genotype-phenotype associations

se.outcome

The standard errors of the genotype-outcome associations

Details

bmi_insulin.

References

Richmond, R. et al., Investigating the role of insulin in increased adiposity: Bi-directional Mendelian randomization study. bioRxiv, 2017, doi:10.1101/155739.

---

Dataset from Do et al., Nat Gen, 2013 containing summary level data on associations of genotypes with lipid traits and the risk of coronary heart diseases

Description

A summary-level dataset, from Do et al. (2013) doi:10.1038/ng.2795, containing 185 single nucleodtide polymorphisms (SNPs) which have genotype-phenotype associations and standard errors for LDL-C, HDL-C, Triglycerides, and genotype-outcome associations for coronary heart disease with their respective standard errors.

Usage

dodata

Format

A data frame with 185 rows and 21 columns with the following variables:

rsid

RSID number

a1

Allele 1

a2

Allele 2

chr

Chromosome

pos

Genomic position

ldlcbeta

The genotype-low-density lipoprotein cholestrol associations

hdlcbeta

The genotype-high-density lipoprotein cholestrol associations

tgbeta

The genotype-triglyceride associations

chdbeta

The genotype-coronary heart disease associations, on the log odds ratio scale

ldlcp2

P-value for genotype-LDL-C associations

hdlcp2

P-value for genotype-HDL-C associations

tgp2

P-value for genotype-triglyceride associations

chdp2

P-value for genotype-coronary heart disease associations

ldlcz

Z-score for genotype-LDL-C associations

ldlcse

The standard errors of the genotype-low-density lipoprotein cholestrol associations

hdlcz

Z-score for genotype-HDL-C associations

hdlcse

The standard errors of the genotype-high-density lipoprotein cholestrol associations

tgz

Z-score for genotype-triglyceride associations

tgse

The standard errors of the genotype-triglyceride cholestrol associations

chdz

Z-score for genotype-coronary heart disease associations

chdse

The standard errors of the genotype-coronary heart disease associations

Details

dodata.

References

Do, R. et al., Common variants associated with plasma triglycerides and risk for coronary artery disease. Nature Genetics, 2013, 45, 1345-1352, doi:10.1038/ng.2795.

---

Bayesian implementation of the MR-Egger multivariate model with choice of prior distributions fitted using JAGS.

Description

Bayesian implementation of the MR-Egger multivariate model with choice of prior distributions fitted using JAGS.

Usage

mr_egger_rjags(
  object,
  prior = "default",
  betaprior = "",
  sigmaprior = "",
  n.chains = 3,
  n.burn = 1000,
  n.iter = 5000,
  seed = NULL,
  rho = 0.5,
  ...
)

Arguments

object	A data object of class mr_format.
prior	A character string for selecting the prior distributions; "default" selects a non-informative set of priors; "weak" selects weakly informative priors; "pseudo" selects a pseudo-horseshoe prior on the causal effect; "joint" selects a joint prior on the intercept and slope.
betaprior	A character string in JAGS syntax to allow a user defined prior for the causal effect.
sigmaprior	A character string in JAGS syntax to allow a user defined prior for the residual standard deviation.
n.chains	Numeric indicating the number of chains used in the MCMC estimation, the default is 3 chains.
n.burn	Numeric indicating the burn-in period of the Bayesian MCMC estimation. The default is 1000 samples.
n.iter	Numeric indicating the number of iterations in the Bayesian MCMC estimation. The default is 5000 iterations.
seed	Numeric indicating the random number seed. The default is the rjags default.
rho	Numeric indicating the correlation coefficient input into the joint prior distribution. The default value is 0.5.
...	Additional arguments passed through to rjags::jags.model().

Value

An object of class eggerjags containing the following components:

AvgPleio

The mean of the simulated pleiotropic effect

CausalEffect

The mean of the simulated causal effect

StandardError

Standard deviation of the simulated causal effect

sigma

The value of the residual standard deviation

CredibleInterval

The credible interval for the causal effect, which includes the lower (2.5%), median (50%) and upper intervals (97.5%)

samples

Output of the Bayesian MCMC samples

Priors

The specified priors

References

Bowden et. al., Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. International Journal of Epidemiology 2015. 44(2): p. 512-525. doi:10.1093/ije/dyv080

Examples

if (requireNamespace("rjags", quietly = TRUE)) {
fit <- mr_egger_rjags(bmi_insulin)
summary(fit)
plot(fit$samples)
# 90% credible interval
fitdf <- do.call(rbind.data.frame, fit$samples)
cri90 <- sapply(fitdf, quantile, probs = c(0.05, 0.95))
print(cri90)
}

---

Bayesian inverse variance weighted model with a choice of prior distributions fitted using Stan

Description

Bayesian inverse variance weighted model with a choice of prior distributions fitted using Stan.

Usage

mr_egger_stan(
  data,
  prior = 1,
  n.chains = 3,
  n.burn = 1000,
  n.iter = 5000,
  seed = 12345,
  rho = 0.5,
  ...
)

Arguments

data	A data of class mr_format.
prior	An integer for selecting the prior distributions; 1 selects a non-informative set of priors; 2 selects weakly informative priors; 3 selects a pseudo-horseshoe prior on the causal effect; 4 selects joint prior of the intercept and causal effect estimate.
n.chains	Numeric indicating the number of chains used in the HMC estimation in rstan, the default is 3 chains.
n.burn	Numeric indicating the burn-in period of the Bayesian HMC estimation. The default is 1000 samples.
n.iter	Numeric indicating the number of iterations in the Bayesian HMC estimation. The default is 5000 iterations.
seed	Numeric indicating the random number seed. The default is 12345.
rho	Numeric indicating the correlation coefficient input into the joint prior distribution. The default is 0.5.
...	Additional arguments passed through to rstan::sampling().

Value

An object of class rstan::stanfit.

References

Bowden J, Davey Smith G, Burgess S. Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. International Journal of Epidemiology, 2015, 44, 2, 512-525. doi:10.1093/ije/dyv080.

Stan Development Team (2020). "RStan: the R interface to Stan." R package version 2.19.3, https://mc-stan.org/.

Examples

if (requireNamespace("rstan", quietly = TRUE)) {
# Note we recommend setting n.burn and n.iter to larger values
suppressWarnings(egger_fit <- mr_egger_stan(bmi_insulin, n.burn = 500, n.iter = 1000, refresh = 0L))
print(egger_fit)
}

---

Organises the summary level data for use in the Bayesian MR functions

Description

Organises the summary level data for use in the Bayesian MR functions

Usage

mr_format(rsid, xbeta, ybeta, xse, yse)

Arguments

rsid	A vector of genetic variants used for analysis, if unspecified a vector is automatically generated.
xbeta	A numeric vector of the instrument-phenotype associations.
ybeta	A numeric vector of the instrument-outcome associations.
xse	The standard errors of the instrument-phenotype associations xbeta.
yse	The standard errors of the instrument-outcome associations ybeta.

Value

A formatted data frame for analysis of class mr_format.

Examples

data(bmi_insulin)
dat <- mr_format(
  rsid = bmi_insulin[,"rsid"],
  xbeta = bmi_insulin[,"beta.exposure"],
  ybeta = bmi_insulin[,"beta.outcome"],
  xse = bmi_insulin[,"se.exposure"],
  yse = bmi_insulin[,"se.outcome"]
)
class(dat)

---

Bayesian inverse variance weighted model with a choice of prior distributions fitted using JAGS.

Description

Bayesian inverse variance weighted model with a choice of prior distributions fitted using JAGS.

Usage

mr_ivw_rjags(
  object,
  prior = "default",
  betaprior = "",
  n.chains = 3,
  n.burn = 1000,
  n.iter = 5000,
  seed = NULL,
  ...
)

Arguments

object	A data object of class mr_format.
prior	A character string for selecting the prior distributions; "default" selects a non-informative set of priors; "weak" selects weakly informative priors; "pseudo" selects a pseudo-horseshoe prior on the causal effect.
betaprior	A character string in JAGS syntax to allow a user defined prior for the causal effect.
n.chains	Numeric indicating the number of chains used in the MCMC estimation, the default is 3 chains.
n.burn	Numeric indicating the burn-in period of the Bayesian MCMC estimation. The default is 1000 samples.
n.iter	Numeric indicating the number of iterations in the Bayesian MCMC estimation. The default is 5000 iterations.
seed	Numeric indicating the random number seed. The default is the rjags default.
...	Additional arguments passed through to rjags::jags.model().

Value

An object of class ivwjags containing the following components:

CausalEffect

The mean of the simulated causal effects

StandardError

Standard deviation of the simulated causal effects

CredibleInterval

The credible interval for the causal effect, which indicates the lower (2.5%), median (50%) and upper intervals (97.5%)

samples

Output of the Bayesian MCMC samples with the different chains

Priors

The specified priors

References

Burgess, S., Butterworth, A., Thompson S.G. Mendelian randomization analysis with multiple genetic variants using summarized data. Genetic Epidemiology, 2013, 37, 7, 658-665 doi:10.1002/gepi.21758.

Examples

if (requireNamespace("rjags", quietly = TRUE)) {
fit <- mr_ivw_rjags(bmi_insulin)
print(fit)
summary(fit)
plot(fit$samples)
# 90% credible interval
fitdf <- do.call(rbind.data.frame, fit$samples)
cri90 <- quantile(fitdf$Estimate, probs = c(0.05, 0.95))
print(cri90)
}

---

Bayesian inverse variance weighted model with a choice of prior distributions fitted using RStan.

Description

Bayesian inverse variance weighted model with a choice of prior distributions fitted using RStan.

Usage

mr_ivw_stan(
  data,
  prior = 1,
  n.chains = 3,
  n.burn = 1000,
  n.iter = 5000,
  seed = 12345,
  ...
)

Arguments

data	A data of class mr_format.
prior	An integer for selecting the prior distributions; 1 selects a non-informative set of priors; 2 selects weakly informative priors; 3 selects a pseudo-horseshoe prior on the causal effect.
n.chains	Numeric indicating the number of chains used in the HMC estimation in rstan, the default is 3 chains.
n.burn	Numeric indicating the burn-in period of the Bayesian HMC estimation. The default is 1000 samples.
n.iter	Numeric indicating the number of iterations in the Bayesian MCMC estimation. The default is 5000 iterations.
seed	Numeric indicating the random number seed. The default is 12345.
...	Additional arguments passed through to rstan::sampling().

Value

An object of class rstan::stanfit.

References

Burgess, S., Butterworth, A., Thompson S.G. Mendelian randomization analysis with multiple genetic variants using summarized data. Genetic Epidemiology, 2013, 37, 7, 658-665 doi:10.1002/gepi.21758.

Stan Development Team (2020). "RStan: the R interface to Stan." R package version 2.19.3, https://mc-stan.org/.

Examples

if (requireNamespace("rstan", quietly = TRUE)) {
suppressWarnings(ivw_fit <- mr_ivw_stan(bmi_insulin, refresh = 0L))
print(ivw_fit)
rstan::traceplot(ivw_fit)
}

---

Bayesian radial MR-Egger model with a choice of prior distributions fitted using JAGS.

Description

Bayesian radial MR-Egger model with a choice of prior distributions fitted using JAGS.

Usage

mr_radialegger_rjags(
  object,
  prior = "default",
  betaprior = "",
  sigmaprior = "",
  n.chains = 3,
  n.burn = 1000,
  n.iter = 5000,
  seed = NULL,
  rho = 0.5,
  ...
)

Arguments

object	A data object of class mr_format.
prior	A character string for selecting the prior distributions; "default" selects a non-informative set of priors; "weak" selects weakly informative priors; "pseudo" selects a pseudo-horseshoe prior on the causal effect; "joint" selects a joint prior on the intercept and slope.
betaprior	A character string in JAGS syntax to allow a user defined prior for the causal effect.
sigmaprior	A character string in JAGS syntax to allow a user defined prior for the residual standard deviation.
n.chains	Numeric indicating the number of chains used in the MCMC estimation, the default is 3 chains.
n.burn	Numeric indicating the burn-in period of the Bayesian MCMC estimation. The default is 1000 samples.
n.iter	Numeric indicating the number of iterations in the Bayesian MCMC estimation. The default is 5000 iterations.
seed	Numeric indicating the random number seed. The default is the rjags default.
rho	Numeric indicating the correlation coefficient input into the joint prior distribution. The default is 0.5.
...	Additional arguments passed through to rjags::jags.model().

Value

An object of class radialeggerjags containing the following components:

AvgPleio

The mean of the simulated pleiotropic effect

CausalEffect

The mean of the simulated causal effect

StandardError

Standard deviation of the simulated causal effect

sigma

The mean of the simaulted residual standard deviation

CredibleInterval

The credible interval for the causal effect, which includes the lower (2.5%), median (50%) and upper intervals (97.5%)

samples

Output of the Bayesian MCMC samples

Prior

The specified priors

References

Bowden, J., et al., Improving the visualization, interpretation and analysis of two-sample summary data Mendelian randomization via the Radial plot and Radial regression. International Journal of Epidemiology, 2018. 47(4): p. 1264-1278. doi:10.1093/ije/dyy101.

Examples

if (requireNamespace("rjags", quietly = TRUE)) {
fit <- mr_radialegger_rjags(bmi_insulin)
summary(fit)
plot(fit$samples)
# 90% credible interval
fitdf <- do.call(rbind.data.frame, fit$samples)
cri90 <- quantile(fitdf$Estimate, probs = c(0.05, 0.95))
print(cri90)
}

---

Bayesian inverse variance weighted model with a choice of prior distributions fitted using RStan.

Description

Bayesian inverse variance weighted model with a choice of prior distributions fitted using RStan

Usage

mr_radialegger_stan(
  data,
  prior = 1,
  n.chains = 3,
  n.burn = 1000,
  n.iter = 5000,
  rho = 0.5,
  seed = 12345,
  ...
)

Arguments

data	A data of class mr_format.
prior	An integer for selecting the prior distributions; 1 selects a non-informative set of priors; 2 selects weakly informative priors; 3 selects a pseudo-horseshoe prior on the causal effect; 4 selects joint prior of the intercept and causal effect estimate.
n.chains	Numeric indicating the number of chains used in the HMC estimation in rstan, the default is 3 chains.
n.burn	Numeric indicating the burn-in period of the Bayesian HMC estimation. The default is 1000 samples.
n.iter	Numeric indicating the number of iterations in the Bayesian HMC estimation. The default is 5000 iterations.
rho	Numeric indicating the correlation coefficient input into the joint prior distribution. The default is 0.5.
seed	Numeric indicating the random number seed. The default is 12345.
...	Additional arguments passed through to rstan::sampling().

Value

An object of class rstan::stanfit.

References

Bowden, J., et al., Improving the visualization, interpretation and analysis of two-sample summary data Mendelian randomization via the Radial plot and Radial regression. International Journal of Epidemiology, 2018. 47(4): p. 1264-1278. doi:10.1093/ije/dyy101.

Stan Development Team (2020). "RStan: the R interface to Stan." R package version 2.19.3, https://mc-stan.org/.

Examples

if (requireNamespace("rstan", quietly = TRUE)) {
# Note we recommend setting n.burn and n.iter to larger values
suppressWarnings({
  radegger_fit <- mr_radialegger_stan(bmi_insulin, n.burn = 500, n.iter = 1000, refresh = 0L)
})
print(radegger_fit)
}

---

Convert an object of class MRInput from the MendelianRandomization package to the mrbayes mr_format class

Description

Creates a data.frame with class mr_format from an object of class MRInput generated by MendelianRandomization::mr_input.

Usage

mrinput_mr_format(dat)

Arguments

dat	Object from MendelianRandomization::mr_input.

Value

Object of class mr_format, the mrbayes format

Examples

if (requireNamespace("MendelianRandomization", quietly = TRUE)) {
dat <- MendelianRandomization::mr_input(
  bx = bmi_insulin$beta.exposure,
  bxse = bmi_insulin$se.exposure,
  by = bmi_insulin$beta.outcome,
  byse = bmi_insulin$se.outcome,
  snps = bmi_insulin$rsid
)
dat <- mrinput_mr_format(dat)
head(dat)
class(dat)
}

---

Bayesian implementation of the MVMR-Egger model with choice of prior distributions fitted using JAGS.

Description

Bayesian implementation of the MVMR-Egger model with choice of prior distributions fitted using JAGS.

Usage

mvmr_egger_rjags(
  object,
  prior = "default",
  betaprior = "",
  sigmaprior = "",
  orientate = 1,
  n.chains = 3,
  n.burn = 1000,
  n.iter = 5000,
  seed = NULL,
  rho = 0.5,
  ...
)

Arguments

object	A data object of class mvmr_format.
prior	A character string for selecting the prior distributions; "default" selects a non-informative set of priors; "weak" selects weakly informative priors; "pseudo" selects a pseudo-horseshoe prior on the causal effect;
betaprior	A character string in JAGS syntax to allow a user defined prior for the causal effect.
sigmaprior	A character string in JAGS syntax to allow a user defined prior for the residual standard deviation.
orientate	Numeric value to indicate the oriented exposure
n.chains	Numeric indicating the number of chains used in the MCMC estimation, the default is 3 chains.
n.burn	Numeric indicating the burn-in period of the Bayesian MCMC estimation. The default is 1000 samples.
n.iter	Numeric indicating the number of iterations in the Bayesian MCMC estimation. The default is 5000 iterations.
seed	Numeric indicating the random number seed. The default is the rjags default.
rho	Numeric indicating the correlation coefficient input into the joint prior distribution. The default value is 0.5.
...	Additional arguments passed through to rjags::jags.model().

Value

An object of class mveggerjags containing the following components:

AvgPleio

The mean of the simulated pleiotropic effect

CausalEffect

The mean of the simulated causal effect

StandardError

Standard deviation of the simulated causal effect

sigma

The value of the residual standard deviation

CredibleInterval

The credible interval for the causal effect, which includes the lower (2.5%), median (50%) and upper intervals (97.5%)

samples

Output of the Bayesian MCMC samples

Priors

The specified priors

References

Bowden et. al., Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. International Journal of Epidemiology 2015. 44(2): p. 512-525. doi:10.1093/ije/dyv080

Examples

if (requireNamespace("rjags", quietly = TRUE)) {
## Not run: 
dat <- mvmr_format(
  rsid = dodata$rsid,
  xbeta = cbind(dodata$ldlcbeta,dodata$hdlcbeta,dodata$tgbeta),
  ybeta = dodata$chdbeta,
  xse = cbind(dodata$ldlcse,dodata$hdlcse,dodata$tgse),
  yse = dodata$chdse
)

fit <- mvmr_egger_rjags(dat)
summary(fit)
plot(fit$samples)
# 90% credible interval
fitdf <- do.call(rbind.data.frame, fit$samples)
cri90 <- sapply(fitdf, quantile, probs = c(0.05, 0.95))
print(cri90)

## End(Not run)
}

---

Bayesian implementation of the MVMR-Egger model with choice of prior distributions fitted using RStan.

Description

Bayesian implementation of the MVMR-Egger model with choice of prior distributions fitted using RStan.

Usage

mvmr_egger_stan(
  data,
  prior = 1,
  n.chains = 3,
  n.burn = 1000,
  n.iter = 5000,
  seed = 12345,
  rho = 0.5,
  orientate = 1,
  ...
)

Arguments

data	A data of class mvmr_format.
prior	An integer for selecting the prior distributions; 1 selects a non-informative set of priors; 2 selects weakly informative priors; 3 selects a pseudo-horseshoe prior on the causal effect;
n.chains	Numeric indicating the number of chains used in the HMC estimation in rstan, the default is 3 chains.
n.burn	Numeric indicating the burn-in period of the Bayesian HMC estimation. The default is 1000 samples.
n.iter	Numeric indicating the number of iterations in the Bayesian HMC estimation. The default is 5000 iterations.
seed	Numeric indicating the random number seed. The default is 12345.
rho	Numeric indicating the correlation coefficient input into the joint prior distribution. The default is 0.5.
orientate	Numeric value to indicate the oriented exposure.
...	Additional arguments passed through to rstan::sampling().

Value

An object of class rstan::stanfit.

References

Bowden J, Davey Smith G, Burgess S. Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. International Journal of Epidemiology, 2015, 44, 2, 512-525. doi:10.1093/ije/dyv080.

Stan Development Team (2020). "RStan: the R interface to Stan." R package version 2.19.3, https://mc-stan.org/.

Examples

if (requireNamespace("rstan", quietly = TRUE)) {
# Note we recommend setting n.burn and n.iter to larger values
dat <- mvmr_format(
  rsid = dodata$rsid,
  xbeta = cbind(dodata$ldlcbeta,dodata$hdlcbeta,dodata$tgbeta),
  ybeta = dodata$chdbeta,
  xse = cbind(dodata$ldlcse,dodata$hdlcse,dodata$tgse),
  yse = dodata$chdse
)
suppressWarnings(mvegger_fit <- mvmr_egger_stan(dat, n.burn = 500, n.iter = 1000, refresh = 0L))
print(mvegger_fit)
}

---

Organises the summary level data for use in the Bayesian MR functions

Description

Organises the summary level data for use in the Bayesian MR functions

Usage

mvmr_format(rsid, xbeta, ybeta, xse, yse)

Arguments

rsid	A vector of genetic variants used for analysis, if unspecified a vector is automatically generated.
xbeta	A matrix of multiple instrument-phenotypes associations.
ybeta	A numeric vector of the instrument-outcome associations.
xse	The matrix for corresponding standard errors of the instrument-phenotypes associations xbeta.
yse	The standard errors of the instrument-outcome associations ybeta.

Value

A formatted data frame for analysis of class mvmr_format.

Examples

data(dodata)
dat <- mvmr_format(
  rsid = dodata$rsid,
  xbeta = cbind(dodata$ldlcbeta,dodata$hdlcbeta,dodata$tgbeta),
  ybeta = dodata$chdbeta,
  xse = cbind(dodata$ldlcse,dodata$hdlcse,dodata$tgse),
  yse = dodata$chdse
)
class(dat)

---

Bayesian multivariate inverse variance weighted model with a choice of prior distributions fitted using JAGS.

Description

Bayesian multivariate inverse variance weighted model with a choice of prior distributions fitted using JAGS.

Usage

mvmr_ivw_rjags(
  object,
  prior = "default",
  betaprior = "",
  n.chains = 3,
  n.burn = 1000,
  n.iter = 5000,
  seed = NULL,
  ...
)

Arguments

object	A data object of class mvmr_format.
prior	A character string for selecting the prior distributions; "default" selects a non-informative set of priors; "weak" selects weakly informative priors; "pseudo" selects a pseudo-horseshoe prior on the causal effect.
betaprior	A character string in JAGS syntax to allow a user defined prior for the causal effect.
n.chains	Numeric indicating the number of chains used in the MCMC estimation, the default is 3 chains.
n.burn	Numeric indicating the burn-in period of the Bayesian MCMC estimation. The default is 1000 samples.
n.iter	Numeric indicating the number of iterations in the Bayesian MCMC estimation. The default is 5000 iterations.
seed	Numeric indicating the random number seed. The default is the rjags default.
...	Additional arguments passed through to rjags::jags.model().

Value

An object of class mvivwjags containing the following components:

CausalEffect

The mean of the simulated causal effects

StandardError

Standard deviation of the simulated causal effects

CredibleInterval

The credible interval for the causal effect, which indicates the lower (2.5%), median (50%) and upper intervals (97.5%)

samples

Output of the Bayesian MCMC samples with the different chains

Priors

The specified priors

References

Burgess, S., Butterworth, A., Thompson S.G. Mendelian randomization analysis with multiple genetic variants using summarized data. Genetic Epidemiology, 2013, 37, 7, 658-665 doi:10.1002/gepi.21758.

Examples

if (requireNamespace("rjags", quietly = TRUE)) {
dat <- mvmr_format(
  rsid = dodata$rsid,
  xbeta = cbind(dodata$ldlcbeta,dodata$hdlcbeta,dodata$tgbeta),
  ybeta = dodata$chdbeta,
  xse = cbind(dodata$ldlcse,dodata$hdlcse,dodata$tgse),
  yse = dodata$chdse
)

fit <- mvmr_ivw_rjags(dat)
print(fit)
summary(fit)
plot(fit$samples)
# 90% credible interval
fitdf <- do.call(rbind.data.frame, fit$samples)
cri90 <- sapply(fitdf, quantile, probs = c(0.05, 0.95))
print(cri90)
}

---

Bayesian multivariate inverse variance weighted model with a choice of prior distributions fitted using RStan.

Description

Bayesian multivariate inverse variance weighted model with a choice of prior distributions fitted using RStan.

Usage

mvmr_ivw_stan(
  data,
  prior = 1,
  n.chains = 3,
  n.burn = 1000,
  n.iter = 5000,
  seed = 12345,
  ...
)

Arguments

data	A data of class mvmr_format.
prior	An integer for selecting the prior distributions; 1 selects a non-informative set of priors; 2 selects weakly informative priors; 3 selects a pseudo-horseshoe prior on the causal effect.
n.chains	Numeric indicating the number of chains used in the HMC estimation in rstan, the default is 3 chains.
n.burn	Numeric indicating the burn-in period of the Bayesian HMC estimation. The default is 1000 samples.
n.iter	Numeric indicating the number of iterations in the Bayesian MCMC estimation. The default is 5000 iterations.
seed	Numeric indicating the random number seed. The default is 12345.
...	Additional arguments passed through to rstan::sampling().

Value

An object of class rstan::stanfit.

References

Burgess, S., Butterworth, A., Thompson S.G. Mendelian randomization analysis with multiple genetic variants using summarized data. Genetic Epidemiology, 2013, 37, 7, 658-665 doi:10.1002/gepi.21758.

Stan Development Team (2020). "RStan: the R interface to Stan." R package version 2.19.3, https://mc-stan.org/.

Examples

if (requireNamespace("rstan", quietly = TRUE)) {
dat <- mvmr_format(
  rsid = dodata$rsid,
  xbeta = cbind(dodata$ldlcbeta,dodata$hdlcbeta,dodata$tgbeta),
  ybeta = dodata$chdbeta,
  xse = cbind(dodata$ldlcse,dodata$hdlcse,dodata$tgse),
  yse = dodata$chdse
)
suppressWarnings(mvivw_fit <- mvmr_ivw_stan(dat, refresh = 0L))
print(mvivw_fit)
rstan::traceplot(mvivw_fit)
}

---