Changes in version 0.1.6 (2020-06-11) - Fix: NAMESPACE warning on CRAN. Changes in version 0.1.5 (2020-03-19) - Fix: recon_vcov now allows prior.prec = 0, reconstructing the covariance matrix from a frequentist analysis of the contrast data (or a Bayesian analysis with improper flat priors). - Fix: Bug when excluding treatment 1 from the decision set where incorrect base-case optimal treatment could be chosen. Changes in version 0.1.4 (2019-01-07) - Feature: Diagnostics from recon_vcov when using NNLS are now returned as additional attributes to the matrix (as well as being printed). - Updated references to Phillippo et al. JRSS:A paper, which is now in print. Changes in version 0.1.3 (2018-03-29) - Fix: Previous release introduced a bug where a single column passed to add.columns resulted in an error. - Feature: Warn when arguments passed to nma_thresh are mismatched (nmatype options with X, mu.design, and delta.design), indicating possible user error. Changes in version 0.1.2 (2018-02-15) - Fix: Additional columns specified with add.columns in thresh_forest are now sorted along with the rest of the table when orderby is given. - Fix: Bugs in thresh_forest with infinite or undefined data estimates and variances. - Fix: Covariance matrices created using Matrix package are acceptable always (not just in the block diagonal case). Changes in version 0.1.1 (2018-01-26) - Fix DOI brackets in DESCRIPTION for CRAN. Changes in version 0.1.0 - Feature: Provide additional columns in the table produced by thresh_forest using argument add.columns. - Feature: Thresholds for decisions based on minimal clinically important difference. Some internal API changes required, but these should not affect the end user. Changes in version 0.0.5 - Update citation with DOI - Minor display fix in vignette Changes in version 0.0.4 Version accompanying Phillippo et al. (2017), available from JRSS:A code archive.