Diagnostic and Prognostic Meta-Analysis

Meta-analysis of diagnostic and prognostic modeling studies. Summarize estimates of prognostic factors, diagnostic test accuracy and prediction model performance. Validate, update and combine published prediction models. Develop new prediction models with data from multiple studies.

  • R package available from CRAN and R-Forge.
  • Maintained by Thomas Debray & Valentijn de Jong
  • Data preparation for systematic reviews of prediction model performance via ccalc and oecalc (Debray et al. 2017 and Snell et al. 2017).
  • Meta-analysis of prediction model performance via valmeta (Debray et al. 2017 and Snell et al. 2017).
  • Evaluation of funnel plot asymmetry and publication bias via fat (Debray et al. 2018).
  • Generation of forest plots via forest.

Multivariate Imputation by Chained Equations

Multiple imputation using Fully Conditional Specification (FCS) implemented by the MICE algorithm as described in Van Buuren and Groothuis-Oudshoorn (2011). Each variable has its own imputation model. Built-in imputation models are provided for continuous data (predictive mean matching, normal), binary data (logistic regression), unordered categorical data (polytomous logistic regression) and ordered categorical data (proportional odds). MICE can also impute continuous two-level data (normal model, pan, second-level variables). Passive imputation can be used to maintain consistency between variables. Various diagnostic plots are available to inspect the quality of the imputations.

  • R package available from CRAN and GitHUB (maintained by Stef van Buuren).
  • Imputation of sporadically and systematically missing values in multilevel data via mice.impute.2l.lmer. See Jolani, Debray et al. (2015) and Audigier et al. (2018).

Multiple Imputation by Chained Equations with Multilevel Data

The micemd package provides methods to perform multiple imputation using chained equations in the presence of multilevel data. It includes imputation methods that account for both sporadically and systematically missing values of continuous, binary and count variables. Following the recommendations of Audigier et al. (2018), the choice of the imputation method for each variable can be facilitated by a default choice tuned according to the structure of the incomplete dataset. Allows parallel calculation for 'mice'.

  • R package available from CRAN and GitHUB (maintained by Vincent Audigier).
  • Imputation of sporadically and systematically missing values in multilevel data via mice.impute.2l.2stage.bin (binary data), mice.impute.2l.2stage.norm (continous data) and mice.impute.2l.2stage.pois (count data). See Audigier et al. 2018.
  • Imputation of univariate missing data using a Bayesian generalized linear mixed model with non-informative prior distributions via mice.impute.2l.glm.bin (binary data), mice.impute.2l.glm.norm (continous data) and mice.impute.2l.glm.pois (count data). See Jolani, Debray et al. (2015) and Audigier et al. (2018).
  • Predictive mean matching imputation for multilevel data via mice.impute.2l.2stage.pmm (Audigier et al. 2018).