Changes for Version 4.28.0 (2023-07-21)

• Upgrade pacakge versions used in the tafe conda environment to be consistent with the packages included in the new Anaconda 2023.07 distribution, which includes Python 3.11.

Changes for Version 4.27.0 (2023-04-13)

• Upgrade pacakge versions used in the tafe conda environment to be consistent with the packages included in the new Anaconda 2023.03-0 distribution, which includes Python 3.10.
• Make a number of non-substantive code changes in order to suppress new pylint 2.16 coding-style warnings.

Changes for Version 4.26.0 (2022-10-25)

• Upgrade pacakge versions used in the tafe conda environment to be consistent with the packages included in the new Anaconda 2022.10 distribution.
• Make a number of non-substantive code changes in order to suppress new pylint 2.14 coding-style warnings.

Changes for Version 4.25.0 (2022-09-24)

• Revise Gini coefficient label in the CLI .tab output to clarify that gini coefficient is calculated using just tax returns (rather than the whole population).

Changes for Version 4.24.0 (2022-09-19)

• Enhance the tacopyfiles development tool to revise new text in the model installation documentation.

Changes for Version 4.23.0 (2022-08-15)

• Add pyyaml package to list of packages required for the tafe conda environment.

Changes for Version 4.22.0 (2022-07-13)

• Add GZIP compress utility function.
• Reduce Analyzer.MIN_XS_BIN_OBS from 90 to 55 so that cross-sectional percentile graphs are generated for samples as small as 5,500 tax filing units.

Changes for Version 4.21.0 (2022-05-12)

• Implement the stratified_sample method in the SampleReWeighting class.
• Add optional other-variable adjustment arguments to the MICE class constructor and add associated logic; read the MICE constructor documentation about the new ovar_idx and zero_ovar_below_abs arguments for details on how this new adjustment works.
• Add analyzer module logic that restricts table and graph output to filers if a model has a nonfiler variable.
• Revise setup.py and environment.yml files to require the newer packages included in the just-released 2022.05 Anaconda distribution.
• Add deprecation warning filters to the utils.py module to suppress warnings generated by the new matplotlib 3.5 package.

Changes for Version 4.20.0 (2022-04-11)

• Switch to using maximum likelihood estimation of alpha in the IncomeUnderReporting class synthetic_adjustment_values_infer method. This change improves the statistical precision of the estimate and speeds execution, but produces essentially the same results as does the Monte Carlo estimation technique used previously.
• Improve technique of ensuring alpha estimate is always positive in the IncomeUnderReporting class synthetic_adjustment_values_mean method.
• Expand the discussion of Problem 5 in the documentation of the data preparation tools.

Changes for Version 4.19.0 (2022-04-03)

• Add logic for the SampleReWeighting class uniform_weight_sample method.
• Add ability to avoid duplicate id values when the IncomeUnderReporting class adjust method is adding extra records; also add an optional verbose argument to the adjust method to get details on the sample adjustment process.
• Rename the IncomeUnderReporting class synthetic_adjustment_values method as synthetic_adjustment_values_mean in order to clarify its difference from the existing synthetic_adjustment_values_infer method.

Changes for Version 4.18.0 (2022-03-16)

• Enhance the IncomeUnderReporting class to have the ability to adjust survey data in which some high incomes are completely missing (because of non-response by the rich) and other high incomes are under-reported. This adjustment capability uses an approach that is similar, in part, to the approach advocated by Blanchet, Flores, and Morgan (BFM) in their 2018 working paper.
• Add to the documentation of the data-preparation tools an extensive example that explains how to use the enhanced IncomeUnderReporting tool to adjust survey data so it can be used effectively as input to a tax microsimulation model.
• Make several minor enhancements to the log_qq_plot.py utility that compares two distributions by showing tabular results and plotting a quantile-quantile graph of the two distributions of log values.

Changes for Version 4.17.0 (2022-02-27)

• Add to the IncomeUnderReporting class the ability to infer from survey data a Pareto distribution parameter value that can be used to adjust high under-reported incomes in the same survey data. This method is essentially the same as used in a recent OECD working paper.
• Add documentation on the new parameter inference capability in the IncomeUnderReporting class.

Changes for Version 4.16.0 (2022-02-16)

• Add an IncomeUnderReporting class that can adjust under-reported high incomes (typically in survey data) using either actual micro data (typically from tax-return data) or synthetic micro data generated using a few macro assumptions about income under reporting.
• Add documentation on the data-preparation tools recently added to the Framework. The documentation includes tool descriptions as well several examples showing how to use the tools to solve different kinds of data-preparation problems.

Changes for Version 4.15.0 (2022-02-06)

• Add to the Multivariate Imputation by Chained Equations (MICE) class the ability to handle efficiently monotone missing data patterns.
• Add to the Multivariate Imputation by Chained Equations (MICE) class the ability to adjust imputed values in several ways so that missing not at random (MNAR) situations can be handled when the missing data pattern is monotone.
• Upgrade the run specification language parser to one that is built using the newest 1.1.0 version of Lark.

Changes for Version 4.14.0 (2022-01-22)

• Improve the Multivariate Imputation by Chained Equations (MICE) class by streamlining its logic, adding a constructor argument, renaming a constructor argument, and improving its documentation. The basic logic is unchanged, but the logic streamlining causes imputed results to be different from the results produced by the prior 4.13 version even when using the same MICE random number seed.

Changes for Version 4.13.0 (2022-01-17)

• Make the Python packages in the tafe environment be consistent with the packages contained in the latest (2021.11) Anaconda Python distribution, which will require removing the old environment if you already have the Tax-Analyzer-Framework package installed on your computer.
• Add minor improvements to the gini utility function.
• Add a Multivariate Imputation by Chained Equations (MICE) class that imputes values that are missing at random (MAR) or missing completely at random (MCAR). Read the docstring in the new taf/mice.py module for details on how the MICE algorithm has been implemented. Read the taf/dptests/test1.py module for a simple example of using the MICE class in a Python script. For more information on imputing missing data values, consult the van Buuren book.

Changes for Version 4.12.0 (2021-11-08)

• Add a gini utility function that computes a Gini coefficient for weighted data and use that function to report in the CLI/GUI table output Gini coefficients for before-tax and after-tax income for both current law and the reform.

Changes for Version 4.11.0 (2021-10-21)

• Add environment.yml file so that the conda tool can be used to automate the process of installing Anaconda packages required by the Framework and models built with it.
• Revise installation documentation for the Framework and for each Framework-derived model to explain how to create and activate the Tax-Analyzer-Framework environment, tafe.
• Revise documentation for each model built with the Framework to explain how to activate the tafe environment when using or developing the model.

Changes for Version 4.10.0 (2021-10-15)

• Revise installation documentation to indicate that a 64-bit Anaconda Python distribution must be installed in order to use Framework-derived tax microsimulation models.
• Clarify installation documentation given changes in versions 4.9.3 and 4.9.4 that align the minimum required package versions with those included in the current (2021.05) Anaconda Python distribution.
• Improve GUI help on custom tabulation of database dump output by adding instructions on how to start SQLite shell program to execute SQL commands interactively.
• Add two tests of the Framework's policy reform capabilities.

Changes for Version 4.9.4 (2021-09-29)

• Revise minimum required matplotlib version from 3.4.0 (which was released some time ago on March 26, 2021) to 3.3.4 (which is, for some reason, the version still included in the Anaconda distribution as of September 29, 2021).

Changes for Version 4.9.3 (2021-09-29)

• Revise minimum required pandas version from 1.3.0 (which was released some time ago on July 2, 2021) to 1.2.4 (which is, for some reason, the version still included in the Anaconda distribution as of late September, 2021).

Changes for Version 4.9.2 (2021-09-25)

• Fix GUI class so that the list of CSV-formatted input data files, which appears on the Run page, is updated each time the user changes the Project directory.

Changes for Version 4.9.1 (2021-09-20)

• Fix GUI class constructor so that it removes all files ending with .txt or .html from the top-level project directory (so that the newest versions of such help documentation always replace older versions).
• Fix the Growth class constructor so that any CSV-formatted file containing growth factors can be used by the class.
• Fix logic of GUI _parameter_value_is_valid function so that value expressions containing -. are invalid (because they are illegal JSON).

Changes for Version 4.9.0 (2021-09-05)

• Update required Anaconda package versions in setup.py and in www/anaconda.md documentation.
• Call pandas DataFrame sum method with numeric_only=True argument in order to avoid deprecation warning introduced when upgrading from pandas 1.2 to 1.3.
• Make non-substantive changes in code and add pylint-disable comments in order to suppress pylint warnings that appear when upgrading from pylint 2.7 to 2.9.

Changes for Version 4.8.0 (2021-08-30)

• Add to plot utility the ability to generate a cross-section after-tax income graph that displays data for a second alternative reform.
• Add --extract option to the csvshow tool.
• Revise the tamakefast utility so that it uses numba.njit parallel=True and prange in calcfuncsfast.py functions that contain a loop through Records variables, a change that speeds up tax calculations by using a thread for each CPU core.

Changes for Version 4.7.0 (2021-06-14)

• Add aggregate tax change amount at the top of each cross-section distributional graph.
• Add to GUI Output page the ability to conduct custom tabulations, using structured query language (SQL), of dump output written to an SQLite3 database file. Each model comes with at least one such custom SQL script, which can serve as a template for user-specified SQL scripts. Both model-supplied and user-written SQL scripts are available for tabulating dump output from the GUI Output page.
• Add a new GUI Calcopts page that provides the ability to specify options that control the model's mode of calculation.

Changes for Version 4.6.0 (2021-05-22)

• Add to each cross-sectional graph output file an new column that includes the value of income at the bottom of each percentile.
• Revise the plot tool so that it can handle the new format of the cross-sectional graph output files.

Changes for Version 4.5.1 (2021-05-21)

• Fix incorrect revision in Framework's cross-sectional graphing logic introduced in version 4.5.0 to enable cross-sectional graphing in the presence of a bunched income distribution.

Changes for Version 4.5.0 (2021-05-20)

• Add to graphical user interface (GUI) Output page the ability to specify a non-zero minimum percentile for the cross-sectional graphs, which provides a kind of graph zoom capability.
• Revise Framework's cross-sectional graphing logic so that bunched income distributions can be graphed without error.

Changes for Version 4.4.0 (2021-05-16)

• Add to graphical user interface (GUI) the ability to specify a run that generates dump output as either an CSV-formatted file or an SQLite database file.

Changes for Version 4.3.0 (2021-05-10)

• Improve graphical user interface (GUI) in numerous ways.
• Test GUI on Windows.

Changes for Version 4.2.0 (2021-05-02)

• Improve graphical user interface (GUI) in several ways, including allowing GUI quit from any notebook page.

Changes for Version 4.1.0 (2021-04-25)

• Add option to have only start-year weights in the weights file and using a WEIGHTS growth factor to extrapolate the weights to subsequent years.
• Improve graphical user interface (GUI) in several ways, including providing more descriptive names for files available on the Output page.

Changes for Version 4.0.0 (2021-04-14)

• Add graphical user interface (GUI) for all Framework-derived models.
• Simplify specification of behavior response bin edges.

Changes for Version 3.2.0 (2021-03-17)

• Require numba version 0.53.0 or higher in order to allow models built with the Framework to run under Python 3.9.x versions (as well as under Python 3.7.x and Python 3.8.x versions).

Changes for Version 3.1.0 (2021-03-08)

• Make new model development quicker and easier by consolidating definitions of model parameters and variables in a new package.json file, reducing the number of JSON files from sixteen to seven.

Changes for Version 3.0.2 (2021-02-28)

• Fix Analyzer class so that preexisting output files for all years are deleted at the beginning of a model run.
• Fix Analyzer class so that language-words file includes list of variables to which each growth factor is applied.
• Enhance documentation of run specification language and RUNSPEC CLI.

Changes for Version 3.0.1 (2021-02-27)

• Fix run specification language parser and interpreter to handle the "USE DUMPVARS {};" instruction correctly.
• Fix Behavior and Analyzer class logic so that marginal tax rates computed in the behavior response logic are added to dump output even if using minimal dump variables.
• Enhance documentation of run specification language and RUNSPEC CLI.

Changes for Version 3.0.0 (2021-02-22)

• Add a RUNSPEC CLI tool that accepts a single file describing all the details of a model run using a simple run specification language, which is described in the updated telling-the-model-what-to-do documentation for the USW-Tax-Analyzer model, one of the models built using the Framework.
• Remove the JSON CLI --outdir option, but otherwise the JSON CLI works as before.
• Add a second parameter output file that contains just the parameters whose value changes between current-law and the reform policy; this new output file has a .pch extension.

Changes for Version 2.9.0 (2021-01-27)

• Add the taunzip tool that unzips a zip file without decompressing gz-compressed files that are inside the zip file. The taunzip tool has a --password option that allows it to unzip encrypted zip files that are password protected. This tool is helpful to those working on Windows, which does not have this capability.
• Add the jsoncheck tool that checks JSON syntax in a specified file.

Changes for Version 2.8.0 (2021-01-24)

• On Windows add to the tahead, tatail, and tagrep tools logic that expends wildcard characters (* and ?) into a list of file names (so that these tools have on Windows the same capabilities as they do on Linux or on a Mac).

Changes for Version 2.7.0 (2021-01-20)

• Enhance JSON file reading capability so that duplicate keys are detected and generate an error message.
• Minor documentation improvements.

Changes for Version 2.6.0 (2021-01-14)

• Require only pandas version 1.1.3 or higher (because 1.2.0 is not yet available in all Anaconda distributions).

Changes for Version 2.5.0 (2021-01-10)

• Fix typo in screen output generated by the tatools utility.
• Add documentation about SQLite3 as data analysis software.
• Require pandas version 1.2.0 or higher, in order to avoid errors in recent 1.1.? versions.

Changes for Version 2.4.0 (2021-01-07)

• Add CLI --dbdump option that causes the dump output file to be written as an SQLite3 database file (ending in .db instead of .csv) that includes two tables of dumpvars, one table for the reform (named ref) and one table for current-law policy (named clp). The sqlite3 command-line shell, which is included in the Anaconda Python distribution, can be used to do custom tabulations of the dump output database using the SQL declarative table language.
• Add --xmin option to the plot tool that generates standard graphs.
• Drop support for Python 3.6 given that pandas will be doing that in the near future.

Changes for Version 2.3.0 (2020-12-28)

• Add tals CLI tool that displays information on all files and subdirectories in the current working directory.

Changes for Version 2.2.0 (2020-12-27)

• Revise Framework API so that each Framework-derived model has a CLI --xstax output option, where the --xstax output is in a tab-delimited text file containing for each income percentile the reform-induced percentage change in tax liability.
• Revise Framework API so that each Framework-derived model's plot tool can graph the new --xstax output.

Changes for Version 2.1.0 (2020-12-22)

• Revise Framework API to allow, but not require, a Framework-derived model to have a CLI --xsetr output option, where the --xsetr output is in a tab-delimited text file containing for each income percentile the effective tax rate.
• Revise Framework API to include a Plot class, and associated command-line plot tool, that generates graphs (in svg, pdf, or eps format) for output files generated by the CLI --tstax, --xsati, and --xsetr options.

Changes for Version 2.0.0 (2020-12-17)

• Revise Framework API to allow, but not require, a Framework-derived model to have a CLI --tstax output option and/or a CLI --xsati output option, where the --tstax output is in a tab-delimited text file containing time-series aggregate taxes under current-law and the reform for each simulated year and where the --xsati output is in a tab-delimited text file containing for each income percentile the reform-induced percentage change in after-tax income.

Changes for Version 1.15.1 (2020-12-13)

• Fix Analyzer class logic so that matched data are correctly extrapolated to the first tax year.
• Revise CLI --help to clarify that matched data/weights files are assumed to have the same year range as the data/weights files built into the model package.

Changes for Version 1.15.0 (2020-12-05)

• Revise logic so that Framework-derived models have a CLI that automatically looks for a weights file whose name matches the specified INPUT data file name; if there is a match, CLI uses the matching weights file, and the specified growth factors, to extrapolate variables and weights to future simulation years.
• Add more error-checking to the csvjoin tool and fix typos in its --help documentation.

Changes for Version 1.14 (2020-11-26)

• Add quantiles utility function and associated test.

Changes for Version 1.13 (2020-11-24)

• Extensive enhancement of documentation, including most notably the CLI --test command writing language-words and language-rules documentation files that support discussion of the declarative language used to tell the CLI what to do.

Changes for Version 1.12 (2020-11-12)

• Change repository name to Tax-Analyzer-Framework and change name of package from psgtaxanalyzer to taf.
• Revise documentation to reflect name change.

Changes for Version 1.11 (2020-11-02)

• Enhance csvshow CLI tool to display information from a broader range of CSV-formatted files.
• Enhance csvjoin CLI tool so that --var variables (other than those in the --svar list) do not have to be in every --csv file.

Changes for Version 1.10 (2020-10-26)

• Add tacat CLI tool that displays all lines in one or more files.

Changes for Version 1.9 (2020-10-22)

• Add tatool CLI utility that lists all the tools made available by the PSG-Tax-Analyzer framework.
• Generalize code so that behavior response analysis can be conducted for a variable whose name contains one or more underscore characters.

Changes for Version 1.8 (2020-10-16)

• Generalize Policy class constructor to handle the case where its growth argument is a dictionary including raw inflation rates and wage growth rates.
• Add Calculator class methods that return the raw inflation rates and wage growth rates from the Policy object embedded in a Calculator object.

Changes for Version 1.7 (2020-10-11)

• Add FREFORM CLI argument that allows specification in a tax model for a sub-national jurisdiction of a JSON reform file for national jurisdiction tax policy.
• Add flexibility so that each model derived from the framework can include or exclude from the CLI each of the following: the FREFORM argument, the --response option, the --growdiff option, and the --calcopts option.

Changes for Version 1.6 (2020-09-23)

• Add a JSON specification file that includes the names of functions in the calcfunctions.py module that either have a Calculator object as its only argument or return a value that is not a records variable.
• Add ability to require a read-in records variable to be less than or equal to (or greater than or equal to) the value of another read-in records variable.

Changes for Version 1.5 (2020-09-05)

• Make minor code changes to avoid errors arising from upgrade to pylint 2.6 and pandas 1.1; tax results are unchanged.
• Update minimum package versions in documentation.

Changes for Version 1.4 (2020-08-27)

• Fix utility function csv_show so that the csvshow tool can display row content from a wider range of CSV-formatted files.

Changes for Version 1.3 (2020-08-21)

• Revise tests to handle tax data files with names ending in .gz (as well as .zip).
• Revise test of valid values for scale_factor, which is in the records_weights.json file.
• Treat variables with data type numpy.float32 as float variables in the df2csv utility function.
• Add imports of numpy and pandas to automatically-generated calcfuncsfast.py module.

Changes for Version 1.2 (2020-07-22)

• Add CalculatorOptions class and CLI --calcopts to change default values of calculator-options parameters.

Changes for Version 1.1 (2020-07-13)

• Simplify names of CLI output files by eliminating any trailing -# characters from the base output file name.

Changes for Version 1.0 (2020-07-10)

• Add source code and tests that, after nearly a year of private development by the Policy Simulation Group, represent a substantially revised, enhanced, and corrected version of Tax-Calculator, which was originally developed by the Open Source Policy Center.

Changes for Version 0.1 (2020-07-03)

• Add initial versions of documentation files; no source code available.