{smcl} {* *! version 2.0.3 30oct2026}{...} {vieweralsosee "csdid" "help csdid"}{...} {vieweralsosee "csdid postestimation" "help csdid_postestimation"}{...} {vieweralsosee "csdid_estat" "help csdid_estat"}{...} {vieweralsosee "csdid_stats" "help csdid_stats"}{...} {vieweralsosee "csgvar" "help csgvar"}{...} {viewerjumpto "Supported" "csdid_legacy##supported"}{...} {viewerjumpto "Deprecated" "csdid_legacy##deprecated"}{...} {viewerjumpto "Replacements" "csdid_legacy##replacements"}{...} {viewerjumpto "Authors" "csdid_legacy##authors"}{...} {title:Title} {p2colset 5 21 23 2}{...} {p2col:{bf:csdid_legacy} {hline 2}}Utility and deprecated commands carried over from csdid Version 1.82{p_end} {p2colreset}{...} {marker supported}{...} {title:Supported utility commands} {pstd} {bf:csgvar} builds the cohort variable that {cmd:gvar()} expects, from a binary treatment indicator. It is fully supported and has a help topic of its own: see {helpb csgvar}. {p_end} {marker deprecated}{...} {title:Deprecated commands} {pstd} {bf:These four commands are deprecated and will be removed in a future release of csdid.} They ship so that existing do-files keep running, and each prints a notice when invoked. They are unsupported: they are kept only to ease migration away from csdid Version 1.82, and they should not be used in new work. {p_end} {synoptset 18 tabbed}{...} {synopthdr:Command} {synoptline} {synopt :{cmd:csdid_rif}}builds tables from saved RIF variables{p_end} {synopt :{cmd:csdid_table}}formats a results table; an internal helper of csdid Version 1.82{p_end} {synopt :{cmd:dipt}}undocumented utility{p_end} {synopt :{cmd:tsvmat}}creates temporary variables from a matrix{p_end} {synoptline} {pstd} Nothing in {cmd:csdid} calls any of them, so their presence costs nothing: Stata compiles an ado file only when its command is first invoked. {p_end} {pstd} {bf:What the deprecated commands refuse rather than ignore.} Each of them parsed options it then discarded. {cmd:csdid_table} accepted {cmd:level()}, {cmd:noci}, {cmd:cformat()} and {cmd:sformat()} and consulted none of them -- its number formats are fixed and its bounds come from {cmd:e(cband)} at whatever level {cmd:csdid} was run with, so {cmd:csdid_table, level(90)} printed a 90% heading over bounds computed at another level. It now refuses those four with return code 198. {cmd:dipt} parsed {cmd:cluster()} and never passed it on, so it returned unclustered standard errors without saying so; it forwards it now, and its weight specification, which made every weighted call a syntax error, is fixed. {p_end} {marker replacements}{...} {pstd} Two numerical details are frozen at their Version 1.82 values, deliberately: the wild-bootstrap band quantile in {cmd:csdid_rif} selects the order statistic at {it:floor(np)+1}, one position above the type-1 quantile the 2.0.0 engine uses, and {cmd:csdid_table}'s {cmd:p} row is left blank, as the Version 1.82 table always left it. Both reproduce the legacy commands' published behavior, to floating-point rounding: {cmd:csdid_rif} rescales the RIF columns only when an entry is missing, where Version 1.82 always did, so on complete columns the last digits can differ. The 2.0.0 estimation and aggregation commands use the modern definitions. {marker syntax}{...} {title:Syntax and stored results of the deprecated commands} {pstd} Each command keeps its Version 1.82 grammar, stated here so an old do-file can be read without the old manual. {pstd} {cmd:csdid_rif} {it:rifvarlist} {ifin} [{cmd:,} {opt cluster(varname)} {opt level(#)} {opt reps(#)} {opt wboot} {opt seed(#)}]{break} averages the saved RIF columns into coefficients. It is eclass: it posts {cmd:e(b)}, {cmd:e(V)}, {cmd:e(N)}, {cmd:e(sample)}, {cmd:e(cmd)} ({cmd:csdid_rif}), {cmd:e(vcetype)} ({cmd:Robust} or {cmd:WBoot}), {cmd:e(level)}, {cmd:e(clustvar)} and {cmd:e(N_clust)} when clustered, and {cmd:e(cband)} (the wild-bootstrap band matrix) under {cmd:wboot}. {opt level()} must be a confidence level and {opt reps()} a positive count; both are validated before anything is computed or the RNG state moves. {pstd} {cmd:csdid_table} [{cmd:,} {it:frozen options}]{break} redisplays the active {cmd:csdid} or {cmd:csdid_rif} result as the Version 1.82 table and leaves {cmd:r(table)} behind. It is rclass; it posts nothing to {cmd:e()}. The four frozen formatting options ({cmd:level()}, {cmd:noci}, {cmd:cformat()}, {cmd:sformat()}) are refused with return code 198; any other estimator's results are refused with return code 459. {pstd} {cmd:dipt} {it:depvar} [{it:indepvars}] {ifin} {it:weight} [{cmd:,} {opt cluster(varname)} {opt from(matname)}]{break} is a thin shim over {helpb mlexp}; everything it posts is {cmd:mlexp}'s own. {pstd} {cmd:tsvmat} {it:matname}{cmd:,} {opt name(newvarlist)}{break} writes matrix columns into new double variables, expanding the data to the matrix's row count when needed. Every refusal -- an existing or invalid name, a duplicated name, more names than columns -- fires before the dataset changes. It returns nothing. {title:What to use instead} {pstd} {bf:Instead of {cmd:csdid_rif}}, take results as a dataset directly. This is the same information in a form you can table, merge or plot, and it comes from the same estimation path {cmd:csdid} itself reports: {p_end} {phang2}{cmd:. csdid y, ivar(id) time(t) gvar(g)}{p_end} {phang2}{cmd:. estat attgt, saving(results) replace}{p_end} {phang2}{cmd:. use results, clear}{p_end} {pstd} {cmd:estat attgt, saving()} returns one row per estimate with the estimate, its standard error, the test statistic, the p-value and the confidence bounds. Every aggregation exports itself the same way: {cmd:estat event, saving()}, {cmd:estat group, saving()}, {cmd:estat calendar, saving()} or {cmd:estat simple, saving()}. {p_end} {pstd} The saved-RIF workflow itself is still supported for files written by 2.0.0's {opt saverif()}: {cmd:csdid_stats} accepts one with {cmd:csdid_stats using} {it:filename}. Only the table-building command is deprecated. A file written by 2.0.0's {opt saverif()} is not a Version 1.82 RIF file: each column is the centred influence function of its cell, with mean zero, and the cell's ATT(g,t) is kept in the column's {cmd:csdid_attgt} characteristic. {cmd:csdid_rif} therefore refuses such a file; aggregate it with {cmd:csdid_stats using}. A RIF file written by Version 1.82 goes the other way: {cmd:csdid_rif} reads it and {cmd:csdid_stats using} refuses it, so run {cmd:csdid} again with {opt saverif()} to aggregate it with {cmd:csdid_stats}. {cmd:csdid_stats} reads a saved file only through {cmd:using}; loading one with {cmd:use} does not make {cmd:csdid_stats} aggregate it. The Version 1.82 recipe that passed {cmd:csdid_rif} the aggregated columns written by {cmd:csdid_stats, save} does not run, because {cmd:save} is not an option of {cmd:csdid_stats}; see {help csdid##remarks_legacy:Migrating from Stata csdid Version 1.82}. {p_end} {pstd} {bf:Instead of {cmd:csdid_table}}, use the table {cmd:csdid} prints, or {cmd:estat tidy, saving()} when you want the numbers rather than the display. {p_end} {pstd} {bf:{cmd:dipt} and {cmd:tsvmat}} were never part of the documented command surface and have no replacement. {p_end} {marker authors}{...} {title:Authors} {pstd} {cmd:csdid} is by Brantly Callaway, Fernando Rios-Avila, and Pedro H. C. Sant'Anna. Full affiliations and contact addresses, the acknowledgments, how to report a problem, and how to cite the package are in {help csdid##authors:help csdid}. {marker alsosee}{...} {title:Also see} {psee} Online: {helpb csdid}, {helpb csdid_postestimation:csdid postestimation}, {helpb csdid_estat}, {helpb csdid_stats}, {helpb csdid_plot}, {helpb csgvar} {p_end}