- Updates the manual vignette to conform to the accepted JSS manuscript (March 2023)
- Updates the documentation for CRAN release
- Adds the doi:10.18637/jss.v106.i07 for the forthcoming JSS paper to DESCRIPTION and documentation

- Update for the JSS manuscript version dated 30-10-2022
- Hiding knots is now more convenient and automatic by setting the
`hide`

entry in the`brokenstick`

object. This replaces the`whatknots`

argument.

- Adds
`hide`

field to`brokenstick`

object - Adds
`hide`

arguments to`coef.brokenstick()`

,`summary.brokenstick()`

,`plot()`

,`get_knots()`

and`get_omega`

- Adds
`cor`

and`lower`

arguments to`summary.brokenstick()`

- Replaces
`what`

argument of`get_omega()`

by`cor`

- Deprecates
`whatknots`

argument in`plot()`

,`get_knots()`

and`get_omega()`

- Separates
`summary()`

and`print()`

functionality - Updates
`smocc_200`

and`fit_200`

objects - Updates to
`roxygen 7.2.1`

- Replace hard-coded variable name
`hgt_z`

by a dynamic name (#8)

- Replaces
`what`

argument in`get_knots()`

to`whatknots`

- Adds support for
`whatknots`

in`get_omega()`

- Extends capabilities of
`plot_trajectory()`

with`shape`

and`linetype`

options - Adds an example to
`plot.brokenstick()`

on how to create a decent black and white figure of trajectories - Replaces
`knots = 0:3`

by`knots = 0:2`

in examples - Updates the perfectmodel vignette

- Changes the default number of knots in
`brokenstick()`

to 5. The former default produced a solution without internal knots. The new default produces a generally more informative starting model when the user does not specify knots (using`knots = c(..., ...)`

) or the number of knots (using`k = ...`

). - Replaces the
`strip_data`

argument in`predict()`

by the a more intuitive`include_data`

argument. By default, observed data are now included into the predictions, similar to`predict.lm()`

. - Turns error
`Argument 'newdata' is required for a light brokenstick object.`

of`brokenstick()`

into a warning and returns`NULL`

. - Updates the vignette Broken Stick Model for Irregular Longitudinal Data to sync with revision for JSS.

- Expression
`predict(fit_200_light, x = "knots")`

now produces warning message instead of crashing - Updates objects
`fit_200`

and`fit_200_light`

to use automatic boundary (2.68y) instead of 3 yrs - Automatically sorts any user-specified values for knots in
increasing order to evade problems with
`predict()`

- Reorganises the vignettes
- Renames
`brokenstick-article.Rmd`

to`manual/manual.Rmd`

, include high-res version on the site and take out of the package to save space - Updates
`vignettes/bibliography.bib`

to title case - Removes superfluous navigation from vigettes
- Defines an
`model.frame.brokenstick()`

function that adheres to conventions - Changes return values by
`fitted()`

and`residuals()`

to vectors - Defines a less verbose
`print.brokenstick()`

helper - Make all calls to
`library()`

to character argument - Removes
`library(lme4)`

from code to evade changing the search path

- Shrinks the size of brokenstick object by removing the
`formula`

list element

- Shrinks the size of light objects by removing the
`sigma2j`

vector from the light`brokenstick`

class

Function

`brokenstick()`

in version`2.0.0`

sets the Kasim-Raudenbush sampler as the default method. The former method`lme4::lmer()`

remains available by setting`method = "lmer"`

argument.Version

`2.0.0`

adopts the variable names of the`coda`

package (e.g.,`start`

,`end`

,`thin`

,`niter`

, and so on) and stores the results of the Kasim-Raudenbush sampler as objects of class`mcmc`

.For

`method = "kr"`

one may now inspect the solution of the sampler by standard functions from the`coda`

package. For`method = "lmer"`

we can apply functions from the`lme4`

package for`merMod`

objects.Version

`2.0.0`

redefines the`brokenstick`

class. New entries include`call`

,`formula`

,`internal`

,`sample`

,`light`

,`data`

,`imp`

and`mod`

. Removed entries are`knots`

(renamed to`internal`

) and`draws`

(renamed to`imp`

). We may omit the`newdata`

argument for the training data. Setting`light = TRUE`

creates a small version of the`brokenstick`

object. Objects of class`brokenstick`

are not backwards compatible, so one should regenerate objects of class`brokenstick`

in order use newer features in`2.0.0`

.Version

`2.0.0`

conforms to classic model fitting interface in`R`

. Renames the`new_data`

argument to`newdata`

to conform to`predict.lm()`

. Methods`plot()`

and`predict()`

no longer require a`newdata`

argument. All special cases of`predict()`

updated and explained in documentation and examples.Version

`2.0.0`

adds methods`coef()`

,`fitted()`

,`model.frame()`

,`model.matrix()`

,`print()`

and`summary`

for the`brokenstick`

object.Simplifies algorithmic control. Renames

`control_brokenstick()`

to`set_control()`

and removes a layer in the control list.

- Stabilises the
`rgamma()`

calls in KR-algorithm for edge cases. `predict_brokenstick()`

can now work with the both (internal) training and (external) test data.- Removes the superfluous
`type`

argument from`predict.brokenstick()`

- Adds a function
`get_omega()`

to extract the variance-covariance matrix of the broken stick estimates - Adds choice
`"dropfirst"`

to`get_knots()`

- Improves error messages of edge cases in
`test-brokenstick_edge.R`

- Perform stricter tests on arguments of
`brokenstick()`

- Introduces argument
`warn_splines`

in`make_basis()`

to suppress uninteresting warns from`splines::bs()`

- Removes superfluous
`knotnames`

argument in`make_basis()`

- Argument
`x`

in`make_basis()`

is now a vector instead of a column vector - Introduces new
`xname`

argument in`make_basis()`

to set the xname

- Handles an edge case that crashed
`predict()`

This version adds a couple of minor alterations.

- Updates cran-comments
- Adds a link to the JSS manuscript in the
`description`

field - Removes unnecessary
`\dontrun{}`

directives - Exports
`parse_formula()`

to remove`:::`

from examples - Sanitises chunk names by removing
`:`

and`_`

characters - Corrects some “first-order” mindo’s to “second-order”
- Repairs plotting glitch in
`oldfriends.Rmd`

- Limits number of printed rows in
`predict.brokenstick()`

example

- Ready for CRAN –> Move up to version 1.0.0
- The package is now hosted on
`https://github.com/growthcharts/brokenstick/`

- Prepare package for first CRAN submission

- Add documentation for the brokenstick class object
- Add JSS manuscript as a vignette
- Remove the prediction vignette and its dependencies
- Extend
`plot.brokenstick()`

with the ability to plot imputed trajectories - Add the
`weightloss`

data - Fail early when user specifies
`degree > 1`

- Adds argument
`what`

to`plot.brokenstick()`

- Solves a bug that always yielded zero rows for case 3 prediction
- Solves a data combination problem in
`predict()`

when the group variable is a factor - Add a better explanation of the
`boundary`

parameter - Evades that
`model.matrix()`

removes rows with`NA`

if`degree = 0`

- This version trims down the package in various ways
- Removes dependencies of
`hardhat`

and`recipes`

- Makes the
`brokenstick`

object smaller since no blueprints are stored - Removes the
`recipe`

interface to the`brokenstick()`

function - Moves
`ggplot2`

to`suggests`

- Copies over the
`install.on.demand()`

function from`mice`

- Imports
`recipes::recipe()`

to inform R package installation process

- Adds badges, resources and references to README
- Updates license
- Updates CITATION

- Removes the dependency on
`growthstandards`

- Updates and corrects
`plot`

examples - Tries to evade
`ggplot2`

out-of-range/missing messages through better filtering

- Adds support for brokenstick model with
`degree = 0`

- This version jump illustrates big and breaking changes:

`brokenstick`

adopted the`tidymodels`

philosophy, and now includes a dependency on`hardhat`

. It is now possible to fit a model using five different interfaces. There is no need anymore the hardcode variable names in the source data.This version introduces a new estimation method, the Kasim-Raudenbush sampler. The new method is more flexible and faster than

`lme4::lmer()`

when the number of knots is large.This version introduces two simple correlation models that may be used to smooth out the variance-covariance matrix of the random effects.

The definition of the

`brokenstick`

class has changed. Objects of class`brokenstick`

do no longer store the training data.The

`brokenstick_export`

class is retired.The

`predict()`

function is fully rewritten as has now a new interface. Since the`brokenstick`

class does not store the training data anymore, the`predict()`

function now obtains a`new_data`

argument. Syntax that worked for`brokenstick`

package before`0.70.0`

does not work anymore and should be updated. The`shape`

argument replaces the`output`

argument.The

`plot()`

function is rewritten, and now requires a`new_data`

specification.Retired functions:

`brokenstick()`

replaces`fit_brokenstick()`

,`predict.brokenstick()`

replaces`predict.brokenstick_export()`

,`get_r2()`

replaces`get_pev()`

Removed functions:

`get_data()`

,`get_X()`

,`export()`

- This version simplifies the plotting functions
- Renders
`ggplot`

objects sharper in vignettes by`svglite`

- Drops the
`pkg`

argument in`plot.brokenstick()`

- Lessens the dependency on
`rbokeh`

- Replaces
`hbgd`

(which is no longer developed) by`growthstandards`

package - Replaces
`smocc_50`

/`fit_50`

by`smocc_200`

/`fit_200`

- Added a
`NEWS.md`

file to track changes to the package

- Added
`smocc_50`

and`fit_50`

demo data - Removed
`smocc.hgtwgt`

,`smocc_hgtwgt`

and`fit_206 datasets`

- Adapt code and vignettes to reflect replacement of demo data

- Adapted source to R 4.0.0.

- Added new utility function
`get_pev()`

for proportion explained variance `get_knots()`

gets a`what`

argument- Now using smarter defaults for
`plot()`

- Simplified arguments to
`plot()`

- Simplified vignette “Overview of main functions”

- Added Support for
`ggplot2`

- Made
`ggplot2`

plot default - Changed default
`show_references`

flag to FALSE

- This is the version announced during my invited lecture at the 7th Channel Network Conference, Hasselt, Belgium.

Here is the abstract of the lecture:

Broken stick model for individual growth curves

Stef van Buuren

- Netherlands Organization for Applied Scientific Research TNO
- Utrecht University

The broken stick model describes a set of individual curves by a linear mixed model using second-order linear B-splines. The model can be used

- to smooth growth curves by a series of connected straight lines;
- to align irregularly observed curves to a common age grid;
- to create synthetic curves at a user-specified set of break ages;
- to estimate the time-to-time correlation matrix;
- to predict future observations.

The user specifies a set of break ages at which the straight lines connect. Each individual obtains an estimate at each break age, so the set of estimates of the individual form a smoothed version of the observed trajectory.

The main assumptions of the broken stick model are that the development between the break ages follows a straight line, and that the broken stick estimates follow a common multivariate normal distribution. In order to conform to the assumption of multivariate normality, the user may fit the broken stick model on suitably transformed data that yield the standard normal (Z-score) scale.

This lecture outlines the model and introduces the brokenstick R package.