v0.7.2

• Fix: Predictions for non-proportional hazards IPD NMA or ML-NMR survival models using aux_regression = ~.trt were incorrectly omitting the treatment effects on the auxiliary parameter(s) in some cases (#43).
• Fix: Calling marginal_effects() for survival outcomes with a single target population previously gave an error.
• Fix: Predictions from exponential models where aux_regression had been specified were giving an error (#44). aux_regression and aux_by have no effect for exponential models since there are no auxiliary (shape) parameters and are ignored, now with a warning.
• Fix: Avoid an error when trying to fit M-spline models combining IPD and AgD in R versions prior to 4.1.0, due to integer coercion of factors by c().

v0.7.1

• Fix: Producing survival/hazard/cumulative hazard predictions for survival models with predict() outside of a plot() call no longer gives an error (#40).
• Fix: Increased StanHeaders version requirement to version 2.32.9 or later, to avoid CRAN sanitizer warnings (caused by stan-dev/rstan#1111).

v0.7.0

• Feature: The new marginal_effects() function produces marginal treatment effects, as a wrapper around absolute predictions from predict(). For example, for an analysis with a binary outcome marginal odds ratios, risk ratios, or risk differences may be produced. For survival outcomes, marginal effects may be based on the full range of predictions produced by predict(), such as marginal differences in restricted mean survival times, or time-varying marginal hazard ratios.
• Feature: Progress bars are now displayed when running interactively for calculations with predict() or marginal_effects() from ML-NMR models that may take longer to run. These can be controlled with the new progress argument.
• Deprecation: The trt_ref argument to predict() has been renamed to baseline_ref; using trt_ref is now soft-deprecated. Renaming this argument baseline_ref follows the naming convention for the other arguments (baseline_type, baseline_level) that specify the details of a provided baseline distribution. This also makes way for the new marginal_effects() functionality.
• Fix: Fallback formatting used by print methods when the crayon package is not installed now works properly, rather than giving errors.
• Fix: Small bug caused predict() for AgD meta-regression models with new data and baseline_type = "response" to fail with an error.
• Fix: The number of studies on a contrast in a network plot plot.nma_data() with weight_edges = TRUE was incorrect when a study had multiple arms of the same treatment. This now correctly counts the number of studies making a comparison, rather than the number of arms.

v0.6.1

• Fix: Piecewise exponential hazard models no longer give errors during set-up. Calculation of RW1 prior weights needed to be handled as a special case.

v0.6.0

Feature: Survival/time-to-event models are now supported

• set_ipd() now has a Surv argument for specifying survival outcomes using survival::Surv(), and a new function set_agd_surv() sets up aggregate data in the form of event/censoring times (e.g. from digitized Kaplan-Meier curves) and overall covariate summaries.
• Left, right, and interval censoring as well as left truncation (delayed entry) are all supported.
• The available likelihoods are Exponential (PH and AFT forms), Weibull (PH and AFT forms), Gompertz, log-Normal, log-Logistic, Gamma, Generalised Gamma, flexible M-splines on the baseline hazard, and piecewise exponential hazards.
• Auxiliary parameters (e.g. shapes, spline coefficients) are always stratified by study to respect randomisation, and may be further stratified by treatment (e.g. to relax the proportional hazards assumption) and/or by additional factors using the aux_by argument to nma().
• A regression model may be defined for the auxiliary parameters using the aux_regression argument to nma(), allowing non-proportionality to be modelled by treatment and/or covariate effects on the shapes or spline coefficients.
• The predict() method produces estimates of survival probabilities, hazards, cumulative hazards, mean survival times, restricted mean survival times, quantiles of the survival time distribution, and median survival times. All of these predictions can be plotted using the plot() method.
• The geom_km() function assists in plotting Kaplan-Meier curves from a network object, for example to overlay these on estimated survival curves. The transform argument can be used to produce log-log plots for assessing the proportional hazards assumption, along with cumulative hazards or log survival curves.
• A new vignette demonstrates ML-NMR survival analysis with an example of progression-free survival after autologous stem cell transplant for newly diagnosed multiple myeloma, with corresponding datasets ndmm_ipd, ndmm_agd, and ndmm_agd_covs.

Feature: Automatic checking of numerical integration for ML-NMR models

• The accuracy of numerical integration for ML-NMR models can now be checked automatically, and is by default. To do so, half of the chains are run with n_int and half with n_int/2 integration points. Any Rhat or effective sample size warnings can then be ascribed to either: non-convergence of the MCMC chains, requiring increased number of iterations iter in nma(), or; insufficient accuracy of numerical integration, requiring increased number of integration points n_int in add_integration(). Descriptive warning messages indicate which is the case.
• This feature is controlled by a new int_check argument to nma(), which is enabled (TRUE) by default.
• Saving thinned cumulative integration points can now be disabled with int_thin = 0, and is now disabled by default. The previous default was int_thin = max(n_int %/% 10, 1).
• Because we can now check sufficient accuracy automatically, the default number of integration points n_int in add_integration() has been lowered to 64. This is still a conservative choice, and will be sufficient in many cases; the previous default of 1000 was excessive.
• As a result, ML-NMR models are now much faster to run by default, both due to lower n_int and disabling saving cumulative integration points.

Other updates

• Feature: dic() now includes an option to use the pV penalty instead of pD.
• Feature: The baseline and aux arguments to predict() can now be specified as the name of a study in the network, to use the parameter estimates from that study for prediction.
• Improvement: predict() will now produce aggregate-level predictions over a sample of individuals in newdata for ML-NMR models (previously newdata had to include integration points).
• Improvement: Compatibility with future rstan versions (PR #25).
• Improvement: Added a plot.mcmc_array() method, as a shortcut for plot(summary(x), ...).
• Fix: In plot.nma_data(), using a custom layout that is not a string (e.g. a data frame of layout coordinates) now works as expected when nudge > 0.
• Fix: Documentation corrections (PR #24).
• Fix: Added missing as.tibble.stan_nma() and as_tibble.stan_nma() methods, to complement the existing as.data.frame.stan_nma().
• Fix: Bug in ordered multinomial models where data in studies with missing categories could be assigned the wrong category (#28).

v0.5.1

This is a bugfix release, primarily to provide compatibility with the latest StanHeaders.

• Fix: Now compatible with latest StanHeaders v2.26.25 (fixes #23)
• Fix: Dealt with various tidyverse deprecations
• Fix: Updated TSD URLs again (thanks to @ndunnewind)

v0.5.0

• Feature: Treatment labels in network plots can now be nudged away from the nodes when weight_nodes = TRUE, using the new nudge argument to plot.nma_data() (#15).
• Feature: The data frame returned by calling as_tibble() or as.data.frame() on an nma_summary object (such as relative effects or predictions) now includes columns for the corresponding treatment (.trt) or contrast (.trta and .trtb), and a .category column may be included for multinomial models. Previously these details were only present as part of the parameter column
• Feature: Added log t prior distribution log_student_t(), which can be used for positive-valued parameters (e.g. heterogeneity variance).
• Improvement: set_agd_contrast() now produces an informative error message when the covariance matrix implied by the se column is not positive definite. Previously this was only checked by Stan after calling the nma() function.
• Improvement: Updated plaque psoriasis ML-NMR vignette to include new analyses, including assessing the assumptions of population adjustment and synthesising multinomial outcomes.
• Improvement: Improved behaviour of the .trtclass special in regression formulas, now main effects of .trtclass are always removed since these are collinear with .trt. This allows expansion of interactions with * to work properly, e.g. ~variable*.trtclass, whereas previously this resulted in an over-parametrised model.
• Fix: CRAN check note for manual HTML5 compatibility.
• Fix: Residual deviance and log likelihood parameters are now named correctly when only contrast-based aggregate data is present (PR #19).

v0.4.2

• Fix: Error in get_nodesplits() when studies have multiple arms of the same treatment.
• Fix: print.nma_data() now prints the repeated arms when studies have multiple arms of the same treatment.
• Fix: CRAN warning regarding invalid img tag height attribute in documentation.

v0.4.1

• Fix: tidyr v1.2.0 breaks ordered multinomial models when some studies do not report all categories (i.e. some multinomial category outcomes are NA in multi()) (PR #11)

v0.4.0

• Feature: Node-splitting models for assessing inconsistency are now available with consistency = "nodesplit" in nma(). Comparisons to split can be chosen using the nodesplit argument, by default all possibly inconsistent comparisons are chosen using get_nodesplits(). Node-splitting results can be summarised with summary.nma_nodesplit() and plotted with plot.nodesplit_summary().
• Feature: The correlation matrix for generating integration points with add_integration() for ML-NMR models is now adjusted to the underlying Gaussian copula, so that the output correlations of the integration points better match the requested input correlations. A new argument cor_adjust controls this behaviour, with options "spearman", "pearson", or "none". Although these correlations typically have little impact on the results, for strict reproducibility the old behaviour from version 0.3.0 and below is available with cor_adjust = "legacy".
• Feature: For random effects models, the predictive distribution of relative/absolute effects in a new study can now be obtained in relative_effects() and predict.stan_nma() respectively, using the new argument predictive_distribution = TRUE.
• Feature: Added option to calculate SUCRA values when summarising the posterior treatment ranks with posterior_ranks() or posterior_rank_probs(), when argument sucra = TRUE.
• Improvement: Factor order is now respected when trt, study, or trt_class are factors, previously the order of levels was reset into natural sort order.
• Improvement: Update package website to Bootstrap 5 with release of pkgdown 2.0.0
• Fix: Model fitting is now robust to non-default settings of options("contrasts").
• Fix: plot.nma_data() no longer gives a ggplot deprecation warning (PR #6).
• Fix: Bug in predict.stan_nma() with a single covariate when newdata is a data.frame (PR #7).
• Fix: Attempting to call predict.stan_nma() on a regression model with only contrast data and no newdata or baseline specified now throws a descriptive error message.