np.gmeta: Make inference for the study-level parameters with no...

In gmeta: Meta-Analysis via a Unified Framework of Confidence Distribution

Description

This method synthesizes information from multiple studies and make inference that is not dependent on any distributional assumption for the study-level parameters. Specifically, the study-level parameters are assumed to be unknown, fixed parameters, it draws inferences about the quantiles of this set of parameters using study-specific summary statistics

Usage

np.gmeta(Thetahat, se, alpha = c(0.025, 0.975), n, m, 
band_pwr = 0.5, resample = 200, B = 40, len = 10)

Arguments

Thetahat	input, point estimate of the true parameter for all K studies. For the kth study, the kth Thetahat is consistent estimator for the true parameter, the kth Theta.
se	input, standard error estimate of th e true parameter estimator for all K studies. For the kth study, there is a standard error estimate, denoted as the kth se, which is associated with the kth Thetahat..
alpha	quantile vector by default.
n	total sample size of all K studies.
m	mth ordered parameter.
band_pwr	a constant in (0,1).
resample	tuning parameter: R realization of CD-random variables for a confidence distribution for the mth ordered parameter.
B	tuning parameter: process with B new "observed" data.
len	tuning parameter: grid search for possible pairs.

Details

The function produces point or quantile estimation for the parameter whether tie or near tie condition exists or not.

Value

An object of class "gmeta.interval", which is a list of following elements:

percentils A three dimensional array containing the min.unif, max.smooth, mean.smooth. min.unif contains quantiles using minimum pair. max.smooth contains quantiles using maximum pair. mean.smooth contains quantiles using mean of all satisfied pairs.

shrink The shrinkage used to obtain presumed "true values".

smoohlist A sequence.

distance The ten by ten dimensional array containing results of loss function.

elig.ind A fourteen by two dimensional array containing counts that how many pairs are smaller than threshold.

Note

Revised on 2017/10/01.

Author(s)

Wei Qian <[e-mail]>, Jerry Q. Cheng <[e-mail]>

References

Brian Claggett, Minge Xie & Lu Tian(2014) Meta-Analysis With Fixed, Unknown, Study-Specific Parameters. Journal of the American Statistical Association, 109 1660-1671.

See Also

gmeta

Examples

 Thetahat = c(-0.102158346,  0.020746333,  0.009118397,  0.163044549, -0.098892204,
              0.161200470,  0.237464018,  0.162000380, -0.088128443,  0.337433537,
              0.032277816,  0.142151631, -0.225430197, -0.057114409,  0.100874063,
              0.177078003, -0.752891602,  0.069141934,  0.114787217, 0.239802656, -0.029858223)
 se = c(0.1742206, 0.1342359, 0.1660455, 0.1372563, 0.1756401, 
 0.1288152, 0.2001390, 0.1603925, 0.1276070, 0.1470815, 
 0.1666182, 0.1694948, 0.1602534, 0.1688223, 0.1512014, 
 0.1857019, 0.1236781, 0.1510481, 0.1733524, 0.1627342, 0.1905231) 
 np.gmeta(Thetahat = Thetahat, se = se, m=10, 
 n =rep(40,21), band_pwr = 0.5, resample=200, B=40, len=10)

gmeta documentation built on March 9, 2021, 9:06 a.m.