It should be noted that Stata's sorting mechanism is not inefficient as a
#XTILE STATA CODE#
Further, compiled C code is much faster than Stata commands. Hence the speed improvements are largest when N / J is Sorting the groups would then be achievable in O(J log(J)) time Sorting a hash can beĪchieved in linear O(N) time, whereas the best sorts take O(N log(N)) Than sorting the data to then process it by group. The key insight is that hashing the data and sorting a hash is a lot faster This does not affect the results in most cases, just the sort order.Ĭommands that take this type of input include: In addition, several commands take gsort-style input, that is As such, theīenchmark is not equivalent and gtoplevelsof does not attempt to implement Has a plethora of features and that one is somewhat incidental. With the goal of gleaning the most common levels of a varlist. It is severalĭozen times faster than 'groups, select', but that command was not written Option, gtoplevelsof does not really have an equivalent. (+) While similar to the user command 'groups' with the 'select' Gquantiles, xtile ( fasterxtile) using by(). (-) fastxtile from egenmisc and astile were benchmarked against Slow due to the way they are written in that case gquantiles is hundreds thousands) pctile and xtile are prohibitively (.) Only gegen group was benchmarked rigorously. (+) The upper end of the speed improvements for gcollapse are for
( NOTE: strL variables are not yet supported see issue 39) Function If you plan to use the pluginĮxtensively, check out the remarks below and the FAQs forĬaveats and details on the plugin (including some extra features!). This package's aim is to provide a fast implementation of various StataĬommands using hashes and C plugins.
#XTILE STATA WINDOWS#
Windows (Cygwin) Faster Stata for Group Operations Levelsof, and unique/distinct using C plugins for a massive speed Implementation of collapse, pctile, xtile, contract, egen, isid, Syntax is "vuong mod1 mod2" where mod1 and mod2 are stored regression results.Faster Stata for big data. See comments in the ado file for syntax.Ĭomputes Vuong (1989 Econometrica) test of two non-nested regressions as implemented and described in Dechow (1994 Journal of Accounting and Economics). Implementation of Mishkin (1983) rational expectations tests. Updated to include option lroc to report area under ROC curve Shumway (2001) hazard model estimates, which uses a standard logit routine and corrects the chi-squared statistics for the average number of observations per cross-sectional unit. OLS regressions with confidence intervals for ratios of regression coefficients based on Fieller's theorem (more robust than delta method) Updated cgmwildboot to fix p-values (e.g., program looks at upper-tail for p-values of positive coefficients, and this was giving p-values greater than one when the coefficient was in the bottom half of the bootstrap distribution) Updated cgmwildboot to fix error introduced by prior fix for dropped variables Updated to deal with dropped variables (e.g., fixed effects dropped due to collinearity) Commands include linear regression (cgmreg), linear regression with Fieller (1954) confidence intervals on coefficient ratios (cgmregF), logit (cgmlogit), fractional logit (cgmflogit) and regression with bootstrapped p-values (cgmwildboot - See Cameron, Gelbach and Miller, 2008 Review of Economics and Statistics). Also see Petersen (2009 Review of Financial Studies). Implementation of various estimation commands with multi-way clustered standard errors as in Cameron, Gelbach and Miller (2010 Journal of Business and Economic Statistics).
#XTILE STATA SERIAL#
Baxter-King (1995) bandpass filter for time series.įama-MacBeth (1973) regressions with options to weight by number of observations as proxy for precision of the years' estimates and an option to use a Newey-West correction for serial correlation in coefficient estimates.
0 kommentar(er)
