logitr (development version)

• Breaking change: the default numDraws for mixed logit models increased from 50 to 500. Draw counts in the tens (the historical convention: {mlogit} and {gmnl} default to 40, Stata’s mixlogit to 50) produce noticeably unstable simulated log-likelihoods, especially with many random parameters – large enough to change model-selection decisions. With the compiled backend the extra cost is a very low overhead on typical models, so a statistically sound default is now affordable. Models estimated with the default will produce (better, but) different results than under previous versions; set numDraws = 50 explicitly to reproduce old results.
• The parallelized multistart now uses the {mirai} package instead of parallel::mclapply()/PSOCK clusters, which gives consistent behavior across platforms (no separate Windows code path). Behavior is otherwise unchanged: the cores are used to run the multistart iterations in parallel, and each model is estimated single-threaded (benchmarking shows this is more efficient than using threads to speed up individual models). The message printed at the start of a multistart now says so. Advanced users can nest the two forms of parallelism by explicitly setting numThreads alongside a multistart. logitr runs its daemons on a dedicated {mirai} compute profile, so it will not interfere with (or tear down) any daemons you have set up yourself.
• Standard deviation estimates for uncorrelated mixed logit models are now always reported as positive. The sign of a standard deviation is not identified (the mixing distributions are symmetric, so sd = s and sd = -s describe the same model), so only the magnitude is meaningful. This avoids the confusion of occasionally seeing a negative standard deviation. The sign change is implemented as an exact relabeling of the converged model (the standard draw columns of flipped parameters are negated along with the coefficient), so the reported log-likelihood, gradient, standard errors, and predictions are exactly those of the converged solution. Correlated models are unaffected, since there the parameters are Cholesky factors whose signs jointly determine the covariance matrix.
• Added a compiled C++ backend for mixed logit models, controlled with the new backend argument. This is now the default for mixed logit (MXL) models (backend = "cpp"), and is roughly 4x faster than the previous R implementation (and faster still with multiple threads; see numThreads) while producing the same results to floating-point precision. Multinomial logit (MNL) models continue to use the R implementation. All mixed logit model types are supported: preference and willingness-to-pay (WTP) space, uncorrelated and correlated heterogeneity, and normal, log-normal, and censored-normal parameter distributions. Set backend = "cpu" to use the native R implementation instead (for example if you need exactly bit-reproducible results). Note that installing the development version from source now requires a C++ compiler (see the installation instructions); the released CRAN version does not.
• Added multithreaded evaluation of the mixed logit log-likelihood via the new numThreads argument (used by backend = "cpp"). The random draws are processed in parallel, giving further speedups on top of the compiled backend that scale with the number of cores. Defaults to using a single thread when running a parallel multistart (numMultiStarts > 1) and all but one of the available cores otherwise.
• Changed the default drawType for mixed logit models from "halton" to "sobol". Sobol draws are better distributed than Halton draws in higher dimensions, where Halton sequences become correlated (roughly when there are more than 5 random parameters). Both are deterministic, so results remain reproducible.
• Added support for Modified Latin Hypercube Sampling draws via drawType = "mlhs" (Hess, Train, and Polak 2006). MLHS draws typically reach a given level of accuracy with fewer draws than Halton draws in higher dimensions. Unlike Halton and Sobol draws, MLHS draws are randomized and are therefore controlled by set.seed().
• Tightened the default optimization stopping tolerances on the objective: ftol_rel and ftol_abs changed from 1e-6 to 1e-10 (the parameter tolerances xtol_rel / xtol_abs are unchanged at 1e-6). With the looser value, runs could stop as soon as the log-likelihood stalled between iterations, sometimes at points with clearly non-zero gradients – which is what occasionally produced NaN standard errors (an indefinite hessian at a non-converged point). The tighter default costs only a few percent more iterations (negligible with the compiled backend) and returns solutions with gradients near zero. Set options = list(ftol_rel = 1e-6, ftol_abs = 1e-6, ...) to reproduce old behavior.
• Fixed the analytic gradient for WTP space models with a censored-normal scale parameter (randScale = "cn"). Draws where the scale censors to zero previously produced NaN entries in the gradient (a 0/0 in the scale parameter’s partials), which caused optimization failures.
• Added the numDrawsBatch argument to stream the random draws in batches when evaluating the mixed logit log-likelihood with backend = "cpu". This keeps peak memory bounded by the batch size rather than the total number of draws, which makes large draw counts (e.g. 10,000+) feasible on the R backend. By default logitr streams automatically only when the draws would otherwise exceed an internal memory budget, so typical models are unaffected. The default "cpp" backend is memory-flat in the number of draws and ignores this argument.

logitr 1.1.2

• Added broom.helpers (>= 1.15.0) and gtsummary (>= 2.0.0) in DESCRIPTION to accord with an update in {gtsummary}.

logitr 1.1.1

• Updated the convergence.Rmd vignette to not run any actual code using other packages to fix issue on CRAN where not all packages are available on all platforms. Now the results are hard-coded in place.
• Fixed bug in makeObsID() where table() was sorting the results stored in reps, which has to be manually undone. Fixes #50.

logitr 1.1.0

• Modified predict() method to use the interval and level arguments of more standard predict() methods.
• Added ci() function.
• Added logit_probs() function.

logitr 1.0.1

• Fixed bug in adjustFactorLevels() where the levels_orig object was being accidentally overwritten.
• Changed the name of adjustFactorLevels() to checkFactorLevels().
• Added a startup message when the package is loaded.

logitr 1.0.0

• Added JSS article DOI throughout package documentation.
• Fixed bug #41 where the predict() method would error if factor levels were missing in newdata.

logitr 0.8.0

• Added tidy(), glance(), and augment() methods for use with the {broom} package.
• Added model.matrix.logitr(), terms.logitr(), and model.frame.logitr() methods for use with {broom.helpers}.
• Added new vignette comparing convergence issues in WTP space models with other similar packages.
• Added new vignette on summarizing results from estimated models.

logitr 0.7.2

• Checks were added to make the panelID and clusterUD variables a sequentially increasing numeric vector and to stop the program if there are any repeated IDs in these variables.
• A patch was added to address a bug in the computation of clustered errors where the data in clusters of size 1 needed to be forced into a matrix with correct dimensions.

logitr 0.7.1

• A patch was added to make the obsID variable a sequentially increasing numeric vector (this was previously done but was accidentally removed in prior updates).
• A patch was added to pass the modelSpace variable along inside the vcov() method.

logitr 0.7.0

• A new vignette on benchmarking was added which tests the package speed against other similar packages.
• A new data set, runtimes, was included, which is exported from the colab notebook used for benchmarking here: https://colab.research.google.com/drive/1vYlBdJd4xCV43UwJ33XXpO3Ys8xWkuxx?usp=sharing
• Sobol draws are supported via a new drawType argument.
• A warning is displayed against using Halton draws after 5 random variables have been specified in a mixed logit model. Users are encouraged to switch to using Sobol draws and increasing the number of draws to at least 200.
• Changed the argument name price to scalePar to be more general.
• Changed the argument name randPrice to randScale to be more general.
• The modelSpace argument is no longer required for specifying a WTP space model as it is redundant. Including a scalePar argument is enough to determine that it is a WTP space model.

logitr 0.6.1

• Fixes error on some of the gradient tests from 0.6.0 release. The gradient tests were sensitive to the random starting parameters, so I modified how the starting parameters were set for the first iteration of a multistart loop such that they would be more consistent.

logitr 0.6.0

• Added the correlation argument to include correlated heterogeneity.
• Added support for new mixed logit distributions: zero-censored normal
• Added new input checks for the obsID and outcome arguments.
• vcov.logitr() method now returns object$vcov if the user set vcov = TRUE during estimation (avoids a redundant calculation of vcov, which is more efficient).
• Added new datasets: apolloModeChoiceData, electricity

logitr 0.5.1

• Patched a bug in how standard errors were being computed when clustering. The source of the error was in re-scaling results post-estimation.

logitr 0.5.0

• The multistart optimization loop is now parallelized.
• Exported the fquantile() function, which is a faster implementation of the stats::quantile() function.

logitr 0.4.0

Larger changes:

• 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

Bugs:

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

logitr 0.3.1

• 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.

logitr 0.3.0

Breaking changes with v0.2.0:

• 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.

Summary of larger updates:

• 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()

Summary of smaller updates:

• 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.

logitr 0.2.7

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

logitr 0.2.6

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.

logitr 0.2.5

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)))

logitr 0.2.4

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.

logitr 0.2.3

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.

logitr 0.2.2

• 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).

logitr 0.2.1

• 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.

logitr 0.2.0

Summary of larger updates:

• 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()

Summary of smaller updates:

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

logitr 0.1.5

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

logitr 0.1.4

• Added predictChoices() function.
• Added predictShares() function, depreciating simulateShares().

logitr 0.1.3

• 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

Bugs

• 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.

logitr 0.1.2

• 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.

logitr 0.1.1

• 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 BX - p. Before it was p + BX 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()

logitr 0.1.0

Summary of larger updates:

• v0.1.0 Submitted to CRAN!

Summary of smaller updates:

• 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".

Bugs

• 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".

logitr 0.0.5

Summary of larger updates:

• Added support for auto creating interactions among variables
• exported getCoefTable() function

Summary of smaller updates:

• 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

logitr 0.0.4

Weighted models, new dataset, new encoding features

Summary of larger updates:

• 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}

Summary of smaller updates:

• 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

logitr 0.0.3

New simulation functionality

Summary of larger updates:

• 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.

Summary of smaller updates:

• 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.

logitr 0.0.2

Updates to options and a few small bug fixes

Summary of larger updates:

• 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.

Smaller updates:

• 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.

Bugs fixed:

• 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.

logitr 0.0.1

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.