| Title: | Bayesian Prior and Posterior Predictive Replication Assessment |
| Version: | 0.1.1 |
| Author: | Yi Zhao [aut, cre], Xiaoquan Wen [aut] |
| Maintainer: | Yi Zhao <zhayi@umich.edu> |
| Description: | Utilize the Bayesian prior and posterior predictive checking approach to provide a statistical assessment of replication success and failure. The package is based on the methods proposed in Zhao,Y., Wen X.(2021) <doi:10.48550/arXiv.2105.03993>. |
| License: | GPL-2 |
| Encoding: | UTF-8 |
| Imports: | mvtnorm, stats, graphics |
| LazyData: | true |
| RoxygenNote: | 7.1.1 |
| Depends: | R (≥ 2.10) |
| NeedsCompilation: | no |
| Packaged: | 2021-12-10 17:20:35 UTC; yizhao |
| Repository: | CRAN |
| Date/Publication: | 2021-12-13 08:20:05 UTC |
Q test statistics
Description
This function calculates the Q test quantity.
Usage
Q(beta, se2, barbeta, phi2, m)
Arguments
beta |
The original or the simulated estimated effects. |
se2 |
The squared standard errors of the estimated effects. |
barbeta |
The estimated true underlying effect. |
phi2 |
The value of the hyperparameter phi. |
m |
The number of replications |
Value
The Q test statistic value
Filtered RPP data
Description
This contains the RP:P data from the Open Science Collaboration project after filtering.
Usage
data("RPP_filtered")
Format
An object of class data.frame with 73 rows and 5 columns.
Examples
data("RPP_filtered")
Egger test statistics
Description
This function provides the calculation for Egger test quantities.
Usage
egger(beta, se2, barbeta, phi2, m)
Arguments
beta |
The original or the simulated estimated effects. |
se2 |
The squared standard errors of the estimated effects. |
barbeta |
The estimated true underlying effect. |
phi2 |
The value of the hyperparameter phi. |
m |
The number of replications |
Value
The egger test statistic value
Cardiovascular disease impact on the mortality of COVID-19
Description
This is a dataset containing several effect estimates and their standard errors for the impact of cardivascular disease on the mortality of COVID-19 in the literature.
Usage
data("mortality")
Format
An object of class data.frame with 6 rows and 3 columns.
Examples
data("mortality")
Posterior Predictive Replication p-value Calculation
Description
Posterior Predictive Replication p-value Calculation
Usage
posterior_prp(
beta,
se,
L = 1000,
r_vec = c(0, 8e-04, 0.006, 0.024),
test = Q,
print_test_dist = FALSE
)
Arguments
beta |
A vector, containing the estimates in the original study and the replication study. |
se |
A vector, containing the standard errors of the estimates in the original study and the replication study. |
L |
A value, determining the times of repeating simulation. |
r_vec |
A vector, defining the prior reproducible model. Each r value corresponds to a probability of sign consistency. |
test |
A function designed to calculate the test quantity, the default one is the Cochran's Q test statistics. |
print_test_dist |
A boolean, determining whether the simulated test statistics value difference will be plot as a histogram or not. Default is False. |
Value
A list with the following components:
grid |
Detailed grid values for the hyperparameters. |
test_statistics |
The test statistics used in calculating the replication p-value. |
n_sim |
The L value. |
test_stats_dif |
The difference between the simulated test statistics quantity and the original value. |
pvalue |
The resulting posterior predictive replicaiton p-value. |
Examples
data("mortality")
res = posterior_prp(beta = mortality$beta, se = mortality$se, test=Q)
names(res)
print(res$pvalue)
Prior Predictive Replication p-value Calculation
Description
Assessing the prior predictive distribution and calculating the replication p-value based on it.
Usage
prior_prp(
beta,
se,
r_vec = c(0, 8e-04, 0.006, 0.024),
test = "two_sided",
report_PI = FALSE
)
Arguments
beta |
A 2-D vector, containing the estimates in the original study and the replication study. |
se |
A 2-D vector, containing the standard errors of the estimates in the original study and the replication study. |
r_vec |
A vector, defining the prior reproducible model. Each r value corresponds to a probability of sign consistency. |
test |
A string, determining which test statistics to utilize. If not specified, the default two-sided one will be used. |
report_PI |
A boolean, denoting whether the 95% predictive interval for the estimates be reported or not. This option is only valid for two-sided test statistics. The default is FALSE. |
Value
A list with the following components:
grid |
The detailed grid values for the hyperparameters. |
test_statistics |
The test statistics used in calculating the replication p-value. |
pvalue |
The resulting prior predictive replicaiton p-value. |
predictive_interval |
The 95% predictive interval if required. |
Examples
data("RPP_filtered")
attach(RPP_filtered)
rpp_pval<-sapply(1:nrow(RPP_filtered),function(x)
prior_prp(beta=c(beta_orig[x], beta_rep[x]),se=c(se_orig[x], se_rep[x]))$pvalue)
Sign consistency probability and the value for r parameter 1-1 transformation
Description
This function transforms the probability of simulated beta_j having the same sign with the underlying true effect barbeta to the corresponding heterogeneity r parameter value.
Usage
prob_to_r(p)
Arguments
p |
A value, the required probability of sign consistency. |
Value
The corresponding heterogeneity parameter value.
Cardiovascular disease impact on the severe case rate of COVID-19
Description
This is a dataset containing several effect estimates and their standard errors for the impact of cardiovascular disease on the severe case rate of COVID-19 in the literature.
Usage
data("severity")
Format
An object of class data.frame with 6 rows and 3 columns.
Examples
data("severity")