Check to see that model object was created using current version of ‘bage’.
Added optimizer argument to fit(), giving choice between three ways of optimizing
Modifed behaviour of quiet argument to fit() so that when it is TRUE, trace output from the optimizer is shown.
Added start_oldpar argument to fit(), to allow calculations to be restarted on a model that has already been fitted.
Modified printing of "bage_mod" object.
bage 0.8.1
Changes to interface
Modified construction of computations part of models so that it works with models fitted using the “inner-outer” method. Extended the print() method for "bage_mod" so that it shows extra output for models fitted using the "inner-outer" method.
bage 0.8.0
Changes to interface
Added ‘along’ column to tidy and print methods for "bage_mod" objects. (Thank you to Andrew Taylor for suggesting this.)
Added zero_sum argument to priors with an along dimension. When zero_sum is TRUE, values for each combination of a by variable and the along variable are constrained to sum to zero. This can allow better identification of higher-level terms in complicated models. It can also slow computations, and has virtually no effect on estimates of the lowest-level rates, probabilities, and means.
Removed post-estimation standardization. We now rely on explicit constraints instead to give interpretable values for main effects and interactions.
Added RW2_Infant() prior for modelling age-patterns of mortality rates.
The s_seas parameter in RW_Seas() and RW2_Seas() now defaults to 0, rather than 1, so that seasonal effects are by default fixed over time rather than varying. Using varying seasonal effects can greatly increase computation times.
Moved rvec from Imports to Depends, so that it loads when bage is loaded. Manipulating results from bage models without rvec loaded can lead to strange errors.
Added information on computations to printout from fitted model objects.
Added function computations(), which can be used to extract this information from fitted model objects.
Added quiet argument to fit(). When quiet is TRUE (the default), warnings generated by nlminb() are suppressed. (These warnings are virtually always about NAs early in the optimization process and are nothing to worry about.)
Changes to internal calculations
Removed some unnecessary coercion of sparse matrices to dense matrices (which could sometimes cause memory problems)
Added extra constraints to some priors - eg the first element of random walks is now zero. This often (but not always) helps make raw estimates of main effects and interactions more interpretable, and can speed up computations slightly.
In the normal model, we now rescale the weights so that they have a mean of 1. This allows us to use the same default prior for dispersion (an exponential prior with mean 1), regardless of the original weights. The rescaling of the weights affects the estimated value for dispersion, but does not affect the estimates for any other parameters.
Generation of posterior sample now using fast methods from package sparseMVN where possible.
bage 0.7.8
Datasets
Added HFD, a scaled SVD object holding data from the Human Fertiltiy Database
Changed names of data objects:
deaths –> isl_deaths
expenditure –> nld_expenditure
divorces –> nzl_divorces
injuries –> nzl_injuries
us_acc_deaths –> usa_deaths
Added new data object kor_births, births in South Korea
bage 0.7.7
Bug fixes
report_sim() now works on fitted models. Thank you to Ollie Pike for pointing out that it previously did not.
Removed redundant levels from age variable in divorces.
Changes to internal calculations
Removed internal bage function rr3(). Call poputils function rr3() instead.
Added minimum version numbers for rvec and poputils.
bage 0.7.5.1
Bug fixes
Fixed bug in code for simulating from Lin() and Lin_AR() priors.
bage 0.7.5
Changes to interface
Added arguments method and vars_inner to fit(). When method is "standard" (the default) fit() uses the existing calculation methods. When method is "inner-outer", fit() uses a new, somewhat experimental calculation method that involves fitting an inner model using a subset of variables, and then an outer model using the remaining variables. With big datasets, "inner-outer" can be faster, and use less memory, but give very similar results.
Added information on numbers of parameters, and standard deviations to output for print. Thank you to Duncan Elliot for suggesting printing numbers of parameters.
Changes to calculations
fit() now internally aggregates input data before fitting, so that cells with the same combinations of predictor variables are combined. This increases speed and reduces memory usage.
Changes to documentation
Added help for print.bage_mod
bage 0.7.4
CRAN release: 2024-08-28
Changes to interface
Function ssvd() no longer exported. Will export once package bssvd matures.
Added first data model. New function is set_datamod_outcome_rr3(), which deals with the case where the outcome variable has been randomly rounded to base 3.
augment() now creates a new version of the outcome variable if (i) the outcome variable has NAs, or (ii) a data model is being applied to the outcome variable. The name of the new variable is created by added a . to the start of the name of the outcome variable.
A help page summarising available data models
bage 0.7.1
Changes to interface
There are now three choices for the standardization argument: "terms", "anova", and "none". With "terms", all effects, plus assoicated SVD coefficients, and trend, cyclical, and seasonal terms, are centered independently. With "anova", the type of standardization descibed in Section 15.6 of Gelman et al (2014) Bayesian Data Analysis, is applied to the effects.
bage 0.7.0
Changes to calculations
Further simplification of standardization, but likely in future to split into two types of standardization: one that gives an ANOVA-style decomposition of effects, and one that helps with understanding the dynamics of each term.
Changes to infrastructure
Added Makevars file.
Changes to documentation
Stopped referring to second-order walks as equivalent to random walks with drift. (A second-order random walk differs from a random walk in that the implied drift term in a second-order random walk can vary over time.)
bage 0.6.3
Changes to calculations
Changed standardization of forecasts so that forecasts are standardized along the ‘along’ dimension by choosing the values that makes them consistent with time trends in the estimation period, and then standardizing within each value of the along dimensions.
bage 0.6.2
Changes to interface
Removed SVDS(), SVDS_AR(), SVDS_AR1(), SVDS_RW(), and SVDS_RW2() priors. Added indep argument to corresponding SVD priors. SVD priors now choose between ‘total’, ‘independent’ and ‘joint’ models based on (1) the value of indep argument, (2) the value of var_sexgender and the name of the term.
Changes to data
Object HMD now contains 5 components, rather than 10.
bage 0.6.1
Changes to calculations
Fixed problems with standardization of forecast
Added an intercept term to Lin() and LinAR() priors
bage 0.6.0
Issues
Standardization of forecasts not working correctly.
Changed values that are stored in object: removed draws_linpred, added draws_effectfree, draws_spline, and draws_svd. Modified/added downstream functions.
Calculation of ‘along_by’ and ‘agesex’ matrices pushed downwards into lower-level functions.
bage 0.5.1
Changes to interface
Moved HMD code to package bssvd.
bage 0.5.0
Changes to interface
Combined interaction (eg ELin) and main effect (eg Lin) versions of priors
augment() method for bage_mod objects now calculated value for .fitted in cases where the outcome or exposure/size is NA, rather than setting the value of .fitted to NA.
Internal calculations
Standardization of effects only done if components() is called. augment() uses the linear predictor (which does not need standardization.)
Internally, draws for the linear predictor, the hyper-parameters and (if included in model) disp are stored, rather than the full standardized components.
Standardization algorithm repeats up to 100 times, or until all residuals are less than 0.0001.
With the new configuration, calculations for large matrices that previously failed with error message “Internal error: Final residual not 0” are now running.
Simulations
When drawing from the prior, the intercept is always set to 0. Terms with SVD or Known priors are not touched. All other terms are centered.
bage 0.4.0
Changes to back-end for SVD priors
Move most functions for creating ‘bage_ssvd’ objects to package ‘bssvd’.
Allowed number of components of a ‘bage_ssvd’ object to differ from
Bug fixes
Corrected error in calculation of logit in ssvd_comp().