Changelog
MBNMAtime 0.2.4
CRAN release: 20231014
Additions/changes
 Reference SDs can now be used when modelling using SMDs to avoid using studyspecific SDs which can lead to imprecision and heterogeneity.

rank.mbnma()
can now only rank a single parameter (e.g.param
argument must be length 1). This facilitates differentiation between treatment and class parameters.
MBNMAtime 0.2.3
CRAN release: 20230317
Additions/changes
 Truncated priors are used as the default priors for timecourse parameters that can only take positive values (e.g.
et50
) for all functions 
texp()
has been removed titp()
is a more stable parameterisation of this function
MBNMAtime 0.2.2
Additions/changes
 Can now specify numeric values for timecourse parameters in the
method
argument. Can be useful for discrete values that cannot be estimated (e.g. fractional polynomial powers, Hill parameter).  Fractional polynomial powers in
tfpoly()
can only take numeric values from set defined in Jansen 2015.  Integrated TwoComponent Prediction (ITP) function (
titp()
) added 
get.relative()
can be used to combine two MBNMA models to allow different timecourse functions to be fitted to a different set of treatments (see examples in the vignette)  New priors that restrict posterior to positive values where necessary can be easily incorporated.

binplot()
can be used to plot the results of NMAs conducted at multiple time bins. This can be particularly useful to explore which timecourse functions might be appropriate, and to check the validity of MBNMA predictions. 
mb.nodesplit()
can be performed at specific timepoints, in addition to by timecourse parameter 
corparam
set toFALSE
as default
Bug fixes
Minor
 Error with
overlay.nma
argument inplot.mb.predict()
fixed
MBNMAtime 0.2.1
CRAN release: 20210913
Additions/changes

get.relative()
function can be used to calculate relative effects/mean differences between treatments/classes 
cumrank()
added for cumulative ranking plots. Also calculates SUCRA values for each treatment and timecourse parameter at specified followup times (even those at which treatments have not been compared within any study)  Studies reporting change from baseline or absolute means can now be specified in
mb.network()
, or will be automatically inferred from the data (studies with no time=0 are assumed to report change from baseline)  Model intercept (response at time=0) is now conditional on change from baseline for each study

texp()
now implements 2parameter exponential function (though the simpler 1parameter model remains the default)
MBNMAtime 0.2.0
CRAN release: 20210426
Additions/changes
 Added variance adjustment (
covar="varadj"
) for correlation between timepoints  this is now the default inmb.run()
 Added log linear timecourse function (
tloglin()
)  Added spline functions (piecewise linear splines, Bsplines, restricted cubic splines, natural splines)
 Added
overlay.nma
option topredict()
to allow plotting of “lumped” NMA results over MBNMA predictions  Modelling can now incorporate Standardised Mean Differences (
link="smd"
) or Ratios of Means (link="log"
) to allow modelling of studies with different scales 
lower_better
argument used instead ofdecreasing
for rankings  Timecourse functions given to
mb.run()
are now given asclass("timefun")
and timecourse parameters are specified within these functions  Predictions from
predict()
can now be ranked  Forest plots now also plot posterior densities using
ggdist::stat_halfeye()
 Neater outputs when using
print()
orsummary()
 Wishart prior used to model correlations between withinstudy baseline effects on different timecourse parameters, in addition to relative effects on different timecourse parameters.
Bug fixes
Major
 Corrected calculation for Bayesian pvalue in
mb.nodesplit()
MBNMAtime 0.1.3
CRAN release: 20200304
Additions/changes
 Added citation file

plot.mb.network()
now uses alayout
argument that takes an igraph layout function instead oflayout_in_circle
(which was a logical argument). This allows any igraph layout to be plotted rather than just a circle (e.g.igraph::as_star()
)  Objects returned from
plot.mb.network
now have specific igraph attributes assigned to them, which can be easily changed by the user. 
user.fun
now takes a formula as an argument (for example~ (beta.1 * dose) + (beta.2 * dose^2)
) rather than a string. 
mb.network
objects are now stored within lists of most other mb class objects for easy reference of data format
Bug fixes
Major
 Exponential function models were not working previously but the doseresponse function has been rewritten so that it runs the model correctly.
 Ensured comparisons are cycled through correctly in mb.nodesplit
 Ensured
timeplot
raw responses can be plotted by either arm (plotby="arm"
) or relative (plotby="rel"
) effects.