- The multistart optimization loop is now parallelized.
- Exported the
`fquantile()`

function, which is a faster implementation of the`stats::quantile()`

function.

- A new
`predict.logitr()`

method was added for making probability and choice predictions from logitr class objects. - The
`predictProbs()`

and`predictChoices()`

functions were depreciated. - Added new
`fitted.logitr()`

and`residuals.logitr()`

methods. - Added optional
`predict`

argument to the main`logitr()`

function which controls whether predicted probabilities, fitted.values, and residuals are included in the returned object. Default setting is TRUE. - Changed the name of the coefficients vector in the returned object from “coef” to “coefficients” to be consistent with other packages.
- Changed the argument name from “choice” to “outcome” to be more general

- Fixed bug where the returned object contained the scaled data rather than the original, unscaled data

- Bug fix: Cast X object to matrix for single-parameter models
- Updated the logic for clustering with and without panel data
- Added the
`se.logitr()`

method. - Added the
`vcov`

argument to the`logitr()`

function. - Improved vignette on interaction models with individual-specific variable interactions.

- Several arguments were moved out of the previous
`options`

argument and are now passed directly as arguments to`logitr()`

. These include:`numMultiStarts`

,`useAnalyticGrad`

,`scaleInputs`

,`startParBounds`

,`standardDraws`

,`numDraws`

,`startVals`

. The`options`

argument is now only used for options to control the optimization handled by`nloptr()`

. - Options for keeping all model outputs on a multistart were removed.

- Added support for panel data in the log-likelihood function and gradients.
- Several argument names in the
`logitr()`

function were changed to make them easier to understand:`choiceName`

became`choice`

,`obsIDName`

became`obsID`

,`parNames`

became`pars`

,`priceName`

became`price`

,`weightsName`

became`weights`

,`clusterName`

became`cluster`

. If used, old names will be passed to the new argument names and a warning will be displayed. - The log-likelihood and gradient functions were overhauled to improve computational efficiency, resulting in substantially faster estimation for all models.
- The following new methods were introduced:
`print.logitr()`

,`logLik.logitr()`

,`coef.summary.logitr()`

,`vcov.logitr()`

,`terms.logitr()`

- Improved
`summary.logitr()`

and`coef.logitr()`

methods for better printing, now using`printCoefmat()`

. - Added input checks for
`wtp()`

and`wtpCompare()`

functions - Fixed some errors in some of the documentation examples and removed the dontrun commands on all of them.
- Added the
`altIDName`

argument to`predictChoices()`

and`predictProbs()`

to preserve the row order of predictions for each alternative in each set of alternatives. Closes issue #13. - Fixed bug in data encoding where random parameter names were not aligned with encoded data.
- Added input checks for all predict functions.

Added support for panel data in the log-likelihood function and gradients

Major changes were made to the gradient functions, which dramatically improved computational efficiency. MNL and MXL models in either preference or WTP spaces now use the faster implementation of the logit calculations.

This version was the first implementation of an alternative approach for computing the logit probabilities, which increased computational speed. Specifically, the formulation was to compute P = 1 / (1 + sum(exp(V - V_chosen)))

The `vcov()`

method was modified such that it computes the covariance post model estimation. Previously, the covariance matrix was being computed internally in the `logitr()`

function, and `vcov()`

just returned this value, which was computationally much slower.

**Several breaking changes in this version**.

- Several argument names were changed to make them easier to understand. These include:
`choiceName`

–>`choice`

,`obsIDName`

–>`obsID`

,`parNames`

–>`pars`

,`priceName`

–>`price`

,`weightsName`

–>`weights`

,`clusterName`

–>`cluster`

. - Several arguments were moved out of the previous
`options`

argument and are now passed directly as arguments to`logitr()`

. These include:`numMultiStarts`

,`useAnalyticGrad`

,`scaleInputs`

,`startParBounds`

,`standardDraws`

,`numDraws`

,`startVals`

. - Some minor tweaks to printing methods.

- Improved
`summary.logitr()`

and`coef.logitr()`

methods for better printing, using`printCoefmat()`

- Added new methods:
`print.logitr()`

,`logLik.logitr()`

,`coef.summary.logitr()`

,`vcov.logitr()`

- Removed option for keeping all model outputs.
- Added input checks for
`wtp()`

and`wtpCompare()`

functions - Fixed some errors in some of the examples and made them all run (removed dontrun commands).

- Added
`altIDName`

argument to`predictChoices()`

and`predictProbs()`

to preserve the row order of predictions for each alternative in each set of alternatives. Closes issue #13. - Fixed bug in data encoding where random parameter names were not aligned with encoded data.
- Added input checks for all predict functions.

- New prediction functions:
`predictChoices()`

and`predictProbs()`

, and , depreciated`simulateShares()`

. - Added robust covariance matrix calculations.
- Added support for clustering errors.
- Major modifications to the
`recodeData()`

function to improve encoding efficiency. - Depreciated
`dummyCode()`

- Improved documentation across all vignettes for new features.
- Improved explanation of preference space and WTP space utility models in vignettes.

- Added robust covariance matrix calculations.
- Added support for clustering errors.

- Added
`predictChoices()`

function. - Added
`predictShares()`

function, depreciating`simulateShares()`

.

- Modified the
`recodeData()`

and`dummyCode()`

functions for improved speed. - Updated
`simulateShares()`

to work with the automatic dummy coding from the revised`recodeData()`

and`dummyCode()`

functions. - Added support for
`simulateShares()`

to compute shares for multiple sets of alternatives. - Added tests for encoding functions
- Added covariance matrix to model export

- When simulating shares from a WTP model, only accepted a price named “price” rather than something else such as “Price” - fixed this.
- In
`simulateShares()`

, the shares were not correctly computed with a WTP space model because price was still being multiplied by -1. This has been corrected. - Changes to automatic dummy coding were accidentally ignoring factor levels - that’s been fixed.

- Fixed bug where model with single variable would error due to a matrix being converted to a vector in the
`standardDraws()`

function - Fixed bug in
`getCatVarDummyNames()`

- previously used string matching, which can accidentally match with other similarly-named variables. - Fixed bug in
`rowsum()`

where the`reorder`

argument was set to`TRUE`

, which resulted in wrong logit calculations unless the`obsID`

happened to be already sorted.

- Changed how failures to converge are handled. Previously would continue to run a while loop. Now it fails and records the failure, along with appropriate changes in summary() and coef().
- Re-defined the wtp space utility models as B
*X - p. Before it was p + B*X and p was re-defined as -1*p. - If tidyverse library is loaded, data frames were getting converted to tibbles, which broke some things. Fixed this by forcing the input data to be a data.frame()

- v0.1.0 Submitted to CRAN!

- Reduced the length of the title in DESCRIPTION to less than 65 characters.
- Changed package names in title and description to single quotes, e.g: {nloptr} -> ‘nloptr’
- Added reference in description with doi to Train (2009) “Discrete Choice Methods with Simulation, 2nd Edition”.
- Added statements to dummyCode.Rd and statusCodes.Rd
- Added statements to dummyCode.Rd and statusCodes.Rd.
- Updated description for summary.logitr.Rd.
- Modified multiple functions to use message()/warning() instead of print()/cat().
- Added
`algorithm`

to the`options`

input, with the default being set to`"NLOPT_LD_LBFGS"`

.

- Fixed tiny bug in
`getParTypes()`

function - previously was not returning the correct`parNames`

for continuous vs. discrete variables. - Added an input check to make sure the modelSpace argument is either
`"pref"`

or`"wtp"`

. - Added an input check to make sure the
`priceName`

argument is only used when the`modelSpace`

argument is set to`"wtp"`

.

- Added support for auto creating interactions among variables
- exported
`getCoefTable()`

function

- Added new documentation for prepping data:
- overall structure
- dummyCode() function
- handling interactions

- All vignettes proof-read with lots of small changes to examples
- Added a hex sticker

Weighted models, new dataset, new encoding features

- Added support for estimating weighted regressions
- Added and improved documentation for new datasets:
`yogurt`

,`cars_china`

,`cars_us`

- Exported the
`dummyCode()`

function for automatically creating dummy-coded variables in a data frame. - Added support for auto dummy-coding categorical variables prior to model estimation
- Major overhaul of documentation using {pkgdown}

- Changed license to MIT (after doing a bit of reading up on this)
- Fixed dimension-matching issue with user-provided draws for mixed logit models
- Fixed bug in
`modelInputs`

where`obsID`

was not a vector for tibble inputs - Added placeholder hex sticker

New simulation functionality

- Added support for simulating shares for a set of alternatives given an estimated model:
`simulateShares()`

. This is similar to the`predict()`

function in mlogit. - Removed support for using an estimated preference space model as an input in the
`options()`

function. I found this just far too confusing, and instead encourage users to supply a WTP space model with the computed WTP from a preference space model as starting parameters.

- Updated the
`summary()`

and main`logitr()`

functions to keep the basic information (run #, log-likelihood value, number of iterations, and output status) whenever`numMultistarts`

> 1. Previously this information was only kept if`keepAllRuns`

was set to`TRUE`

.

Updates to options and a few small bug fixes

- I got rid of the
`logitr.summary()`

function and instead added the`logitr`

class to all the models and renamed the summary function to`summary.logitr()`

. Now you can just use the standard`summary()`

function to summarize model results. - I finally fixed the analytic gradient for WTP space MXL models. I tested analytic versus numeric for WTP space and Preference Space MXL models and they are all identical, including variations of using normally and log-normally distributed parameters.
- Added startParBounds as an argument in options.

- Changed the summary() function to print more digits in the summary table.
- Rounded printing of the elapsed time in the summary table.
- Forced the sigma values in MXL models to be positive using abs(). Negative values for sigma parameters should not be an issue because the standard normal is symmetric.
- Changed the summary of random parameters to show “summary of 10k draws”
- Updated hessian to always use numeric approx for SE calculation since it’s faster.
- Made scaleInputs default to
`TRUE`

.

- If the prefSpaceModel was a multistart, it was grabbing the correct bestModel for the WTP calculations, but not the logLik value. Now it’s getting the right logLik value too.
- Fixed a bug with the scaling option where it was blowing up to use scaling numbers.

Full reboot of logitr!

Long overdue, I decided to give the logitr program a full overhaul. This is the first version that is compiled as a proper R package that can be directly installed from GitHub. This version is much more robust and flexible than the prior, clunky collection of R files that I had previously been using to estimate logit models.