| Title: | Improved Methods for Constructing Prediction Intervals for Network Meta-Analysis |
|---|---|
| Description: | Improved methods to construct prediction intervals for network meta-analysis. The parametric bootstrap and Kenward-Roger-type adjustment by Noma et al. (2022) <forthcoming> are implementable. |
| Authors: | Hisashi Noma [aut, cre]
|
| Maintainer: | Hisashi Noma <[email protected]> |
| License: | GPL-3 |
| Version: | 1.1-2 |
| Built: | 2024-10-31 18:37:26 UTC |
| Source: | https://github.com/cran/PINMA |
Improved Methods for Constructing Prediction Intervals for Network Meta-Analysis.
Noma, H., Hamura, Y., Sugasawa, S. and Furukawa, T. A. (2022+). Improved methods to construct prediction intervals for network meta-analysis. Forthcoming.
Transforming arm-level data to contrast-based summary statistics.
data.edit(study,trt,d,n)data.edit(study,trt,d,n)
study |
Study ID |
trt |
Numbered treatment (=1,2,...) |
d |
Number of events |
n |
Sample size |
Contrast-based summary statistics are generated.
y: Contrast-based summary estimates.
S: Vectored within-study covariance matrix.
data(dstr) attach(dstr) edat <- data.edit(study,trt,d,n)data(dstr) attach(dstr) edat <- data.edit(study,trt,d,n)
study: Study ID
treat: Treatment
trt: Numbered treatment (1:CCTA, 2:CMR, 3:exercise ECG, 4:SPECT-MPI, 5:standard care, 6:Stress Echo)
n: Sample size
d: Number of events
data(dstr)data(dstr)
A arm-based dataset with 29 rows and 5 variables
Siontis, G. C., Mavridis, D., Greenwood, J. P., et al. (2018). Outcomes of non-invasive diagnostic modalities for the detection of coronary artery disease: network meta-analysis of diagnostic randomised controlled trials. BMJ. 360: k504.
Kenward-Roger-type adjustment for constructing prediction intervals of network meta-analysis.
KR(y, S)KR(y, S)
y |
Contrast-based summary data of the outcome measure |
S |
Covariance estimates of |
Results of the Kenward-Roger-type adjustment for inference of multivariate random-effects model and prediction intervals for network meta-analysis.
Estimates: Restricted maximum likelihood (REML) estimates, their SE, and Wald-type 95% confidence intervals by the Kenward-Roger-type adjustment.
Between-studies_SD: Between-studies SD estimate.
95%PI: 95% prediction intervals by the Kenward-Roger-type adjustment.
Noma, H., Hamura, Y., Sugasawa, S. and Furukawa, T. A. (2022+). Improved methods to construct prediction intervals for network meta-analysis. Forthcoming.
data(dstr) attach(dstr) # Transforming the arm-level data to the contrast-based summaryies edat <- data.edit(study,trt,d,n) y <- edat$y S <- edat$S KR(y,S) # Results of the NMA analysis (log OR scale)data(dstr) attach(dstr) # Transforming the arm-level data to the contrast-based summaryies edat <- data.edit(study,trt,d,n) y <- edat$y S <- edat$S KR(y,S) # Results of the NMA analysis (log OR scale)
Parametric bootstrap procedure for constructing prediction intervals of network meta-analysis.
PBS(y, S, B=2000)PBS(y, S, B=2000)
y |
Contrast-based summary data of the outcome measure |
S |
Covariance estimates of |
B |
Number of bootstrap resampling (default: 2000). |
The parametric bootstrap prediction intervals for network meta-analysis.
Estimates: Restricted maximum likelihood (REML) estimates, their SE, and 95% Wald-type confidence intervals.
Between-studies_SD: Between-studies SD estimate.
95%PI: 95% prediction intervals by the parametric bootstrap.
Noma, H., Hamura, Y., Sugasawa, S. and Furukawa, T. A. (2022+). Improved methods to construct prediction intervals for network meta-analysis. Forthcoming.
data(dstr) attach(dstr) # Transforming the arm-level data to the contrast-based summaryies edat <- data.edit(study,trt,d,n) y <- edat$y S <- edat$S PBS(y,S,B=10) # Results of the NMA analysis (log OR scale); B is recommended to be >= 1000.data(dstr) attach(dstr) # Transforming the arm-level data to the contrast-based summaryies edat <- data.edit(study,trt,d,n) y <- edat$y S <- edat$S PBS(y,S,B=10) # Results of the NMA analysis (log OR scale); B is recommended to be >= 1000.
The ordinary t-approximation for constructing prediction intervals of network meta-analysis.
tPI(y, S)tPI(y, S)
y |
Contrast-based summary data of the outcome measure |
S |
Covariance estimates of |
The ordinary t-approximation prediction intervals for network meta-analysis.
Estimates: Restricted maximum likelihood (REML) estimates, their SE, and Wald-type 95% confidence intervals.
Between-studies_SD: Between-studies SD estimate.
95%PI: 95% prediction intervals by the ordinary t-approximation.
Cooper, H., Hedges, L. V., and Valentine, J. C. (2009). The Handbook of Research Synthesis and Meta-Analysis, 2nd edition. New York: Russell Sage Foundation.
Chaimani, A., and Salanti, G. (2015). Visualizing assumptions and results in network meta-analysis: the network graphs package. Stata Journal 15, 905-920.
data(dstr) attach(dstr) # Transforming the arm-level data to the contrast-based summaryies edat <- data.edit(study,trt,d,n) y <- edat$y S <- edat$S tPI(y,S) # Results of the NMA analysis (log OR scale)data(dstr) attach(dstr) # Transforming the arm-level data to the contrast-based summaryies edat <- data.edit(study,trt,d,n) y <- edat$y S <- edat$S tPI(y,S) # Results of the NMA analysis (log OR scale)