kylehamilton / mavis Goto Github PK

View Code? Open in Web Editor NEW

35.0 7.0 14.0 246 KB

MAVIS: Meta Analysis via Shiny

R 100.00%

meta-analysis shiny cran effect-size

mavis's Introduction

MAVIS: Meta Analysis via Shiny v1.1.4

Live demo of MAVIS v1.1.3 can be found here http://kylehamilton.net/shiny/MAVIS/

June 21, 2016

Examples now include data from published papers
More effect size calculators have been added

August 4, 2015

Version 1.1.1 has been sent to CRAN
Added one new effect size calculator and included the Turkish version of MAVIS (aRma) to the package.

April 29, 2015

Version 1.1 has been sent to CRAN and can be found here https://cran.rstudio.com/web/packages/MAVIS/.

List of Packages Used

library("shiny")
library("shinyAce")
library("shinyBS")
library("metafor")
library("MAd")
library("MAc")
library("quantreg")
library("ggplot2")
library("compute.es")
library("SCMA")
library("SCRT")
library("weightr")
library("irr")

Acknowledgments and Authors

Acknowledgments

W. Kyle Hamilton would like to thank the Health Communications and Interventions Lab at UC Merced for their comments and beta testing efforts on this application as well as Kathleen Coburn for her feedback and evaluation of the statistical methods related to this project.

Atsushi Mizumoto would like to thank Dr. Luke Plonsky and Dr. Yo In'nami for their support and feedback to create this web application.

Authors

W. Kyle Hamilton - University of California, Merced

W. Kyle Hamilton maintains this application and has authored new features.

Burak Aydin, PhD - Recep Tayyip Erdoğan University

Burak Aydin is working on a Turkish version of MAVIS and contributed the dichotomous data entry feature.

Atsushi Mizumoto, PhD - Kansai University

Atsushi Mizumoto wrote the first version of this application; this application is a fork of the original which can be found here

Kathleen Coburn contributed technical advice on how to run a meta-analysis as well as information on publication bias.

mavis's People

Contributors

Stargazers

Watchers

Forkers

1tamas guaponi joscani ericafigbar katiecoburn wcools imanojkumar woodhaha burakaydin ekothe lingtax caroluidhao takewiki chuanchang

mavis's Issues

Provided sample data for correlations are not accurate for the study cited

These correlations are reported to relate to Widman, L., Noar, S. M., Choukas-Bradley, S., & Francis, D. B. (2014). Adolescent Sexual Health Communication and Condom Use : A Meta-Analysis. Health Psychology however there are errors in the reported correlation coefficients for study 24 (Guzmán-2003) and study 27 (Hart-2005).

These studies are entered as 0.334 and 0.07 respectively however both should be negative: -0.334 and -0.07.

Study label bug

Dr. Soliman Said emailed me about a bug where if the first letter for the study label is not capitalized MAVIS will throw an error. For example "Study" will work but "study" will not.

Correlation analysis implemented using meta::metacor does not use raw correlations when raw correlations are selected in "Correlation model measures"

Using the default correlations data from Widman, L., Noar, S. M., Choukas-Bradley, S., & Francis, D. B. (2014). Adolescent Sexual Health Communication and Condom Use : A Meta-Analysis. Health Psychology. results in a RE model estimate of r=0.2650 regardless of whether ZCOR or COR are selected.

The code to implement the FE and RE meta-analysis is FE.res <- meta::metacor(dat$r, dat$N) this does not take an argument related to the model estimator ("DL", "REML" etc) or an argument related to the measure ("ZCOR" or "COR").

meta::metacor(dat_cor$yi, dat_cor$N, method.tau = "DL", sm = "COR") would be required to calculate the estimate using raw correlations.

Fix image output

Image output is not connected

Doesn't show all the results from Meta-analysis

The following data does not generate a Fixed-Effect model forest plot despite it assumes homogeneity.
Study N1 N2 d
Day (2013) 148 148 0.13
Findik (2012) 60 57 0.2
Wassenberg (2010) 42 84 0.01
Lupion (2015) 72 72 0.35
It uses Effect Size d.
Thank's for the software!

Add single case design framework

Show model script?

I was wondering if it would be possible to make the analysis scripts used accessible to the user? In my reviews, I strongly encourage authors to give the analysis scripts for their meta-analyses, and it would be great if I can do that alongside recommending MAVIS.

meta dependence

When I installed and ran MAVIS I got an error message:

library(shiny)
library(MAVIS)
startmavis()

Listening on http://127.0.0.1:7319
Warning: Error in library: there is no package called ‘meta’
Stack trace (innermost first):
42: library
2: shiny::runApp
1: startmavis
Error : there is no package called ‘meta’

It is probably some issue with the dependencies of packages because when I manually installed meta package it worked fine again.