Missing Data and Prediction. R packages psfmi and miceafter.
Joined January 2015
- Tweets130
- Following179
- Followers140
- Likes1.3K
Martijn Heymans retweeted
In AJE:
Missing data? Don't drop observations! A longitudinal study in @hrsisr shows multiple imputation with predictive mean matching performs well, is easy to implement, and helps recover unbiased results
doi.org/10.1093/aje/kwad139
Just released on cran, psfmi package version 1.4.0 including new possibility to pool and select stratified Cox models after Multiple Imputation #rstats mwheymans.github.io/psfmi/
Martijn Heymans retweeted
@RoelWing just published his last PhD paper in the @JPhysiother!
A prognostic model for neck-related disability in patients with sub-acute neck pain was externally validated at 6 weeks with acceptable discrimination and calibration.
sciencedirect.com/science/ar…
Finding it hard to understand the Cox model and baseline hazard in R. See my post that may help you!
missingdatasolutions.rbind.i… #RStats #rstudio
Martijn Heymans retweeted
His publications are suggested reading for anyone interested in prognostic modelling for neck pain.
Starting with this systematic review...
pubmed.ncbi.nlm.nih.gov/2928…
...to this paper failing to externally validate the most promising models:
pubmed.ncbi.nlm.nih.gov/3430…
2/3
Martijn Heymans retweeted
A systematic review shows no performance benefit of machine learning over logistic regression for clinical prediction models jclinepi.com/article/S0895-4…
Proud to present this paper about Handling Missing Data in Clinical Research as a Key Concept in Clinical Epidemiology @JClinEpi eur04.safelinks.protection.o…
Martijn Heymans retweeted
Ever wondered whether it is necessary to impute cost data before using longitudinal linear mixed-model analyses? Check out our paper below for the answer! @hannekevandonge @VroomenJanet @mwheymans doi.org/10.1007/s10198-022-0…
Great paper with a great team! ejnmmires.springeropen.com/a…
Martijn Heymans retweeted
#openaccess Time lags and time interactions in mixed-effects models impacted longitudinal mediation effect estimates jclinepi.com/article/S0895-4…
Do not dichotomize variables! Read here why, including easy to understand explanation and interpretation of linear and restricted cubic splines + R code.
frontiersin.org/articles/10.…
Excited to share this NEW PAPER about pooling and selecting prognostic/prediction models after Multiple Imputation using the median of p-values rule (MPR) that outperforms more complex methods. eur04.safelinks.protection.o…
Martijn Heymans retweeted
We are excited to announce that RStudio, PBC will become Posit, PBC in October!
More information here: rstudio.com/blog/rstudio-is-…
Martijn Heymans retweeted
#openaccess Missing data were poorly reported and handled in randomized controlled trials with repeatedly measured continuous outcomes: a cross-sectional survey jclinepi.com/article/S0895-4…
Martijn Heymans retweeted
Reminder: Tomorrow is the follow-up session to the 'Meta-Analysis with R' workshop I gave prior to #ESMARConf2022. Details here: wvbauer.com/doku.php/worksho… We will continue where I left off (see here for the materials covered so far: github.com/wviechtb/esmarcon…).
Want to learn about the added value of splines? See: Proposed New Dynamic Prognostic Index for Diffuse Large B-Cell Lymphoma: International Metabolic Prognostic Index pubmed.ncbi.nlm.nih.gov/3535…
Martijn Heymans retweeted
Missing data was handled inconsistently in UK prediction models: a review of method used jclinepi.com/article/S0895-4…
Martijn Heymans retweeted
Attending the online @APH Methodology tutorial by @JudithRijnhart: about the robustness of #multiverse analysis - assessing the effect of the researcher's degrees of freedom on the #replicability of an analysis.
with @judithbosmans
link.springer.com/article/10…
