{smcl}
{* *!version 1.0 31/1/2024}{...}
{cmd:help dltable}
{hline}
{title:Title}

{phang}
{bf:dltable} {hline 2}  Regression tables for Randomized Controlled Trials Using Double LASSO 




{title:Syntax}
{p 8 17 3}
{cmdab:dltable}
 {depvar} 
[if]
{weight} 
{cmd:,} {it:options}


{synoptset 20 tabbed}{...}
{synopthdr}
{synoptline}
{syntab:Main}
{synopt:{opth treat:ment(treatvar)}} specify the treatment variable to be used. Multiple branches are possible. In that case, a dummy variable for each treatment branch should be included i.e. each dummy should take value 1 if treatment and 0 otherwise. Do not include a dummy for the control group. Factor variables cannot be used. Hterogeneous analysis is possible if you include the interaction terms in the treatment options (as if they were additional treatment branches)   {p_end} 
{synopt:{    opth  clust:er(varname)}}    specify the clustering variable if any {p_end} 
{synopt:{ opth  control(varlist)}}   specify the variables used for the double LASSO estimation. These variables are used for every single outcome variables. You may use factor variable (fixed effects) but only if in numeric format. {p_end} 
{synopt:{ opth  partialout(varlist)}}   specify a set of control variables to be used in the estimation. partialout option forces these variables to be included in the double LASSO algorithmThese variables are used for every single outcome variables. You may use factor variable (fixed effects) but only if in numeric format.  {p_end} 
{synopt:{ opt  keep }} instead of dropping the columns after running, the program keeps them in the dataset. {p_end} 
{synopt:{ opt  sd }} display the standard deviations of the average and of the control group below in square parenthesis.  
{p_end}
{synopt:{ opt  quiet }} do not show the regression outcomes. {p_end} 
{synopt:{ opt  sheet(string asis[, replace|modify])}} replace a worksheet in an already existing excel document. Sub-options replace or modify are necessary if the worksheet already exists (see export excel command) {p_end} 
{synopt:{ opt  qval(name)}} display the qvalues in square parenthesis underneath the standard error.  The qvalues are computed using the qqvalue command and the fdr sharpened qvalue strategy by Anderson(2008).   
Varname specifies the multiple-test procedure method to be used for calculating the q-values from the input P-values. The method is one of bonferroni | sidak | holm | holland | hochberg | simes | yekutieli | bky.  These method names specify that the q-values will be calculated from the input P-values by inverting the multiple-test procedure specified by the method() option of the same name for the multproc option of the smileplot package. 
bky method does not rely on qqvalue command but corresponds to the sharpened qvalue developped by Benjamini etal. (2006) and described in Anderson(2008). Command qqvalue is automatically installed from SSC when option qvalue is specified. {p_end}
{synopt:{ opt  selection(name)}} is the selection option from the dsregress command. See dsregress for details. By default, dltable uses the plugin algorithm.  {p_end}
{synoptline}

{p2colreset}{...}
{p 4 6 2}
{cmd:if} is allowed.{p_end}
{p 4 6 2}
{cmd:fweight, aweight, pweight}s are allowed; see {help weight}. {p_end} 

{title:Description}

{pstd}
 {cmd:dltable } creates regressions and tables (with the subcommand using) for experimental studies using double LASSO estimation (Belloni et al., 2014) It is the sister command of rctable. dltable is particularly well adadpted to Randomized Controlled Trial or any analysis comparing different (experimental) groups.  dltable creates several variables: VAR (the variable name) LAB (its label) A (the average in both groups) C (the average in the control group), and COEF1 [COEF2...], which record the treatment control different estimated via double LASSO. With the using option, dltable exports the table in an excel, csv or text format. The variables created are then 
dropped, unless the keep option is specify. dltable can also be used without using.  The command accepts multiple treatment branches and can also be used to create tables with interaction terms. For multiple treatment branches, include one dummy per treatment branch but do not include a control dummy.  The double LASSO estimation is conducted using dsregress stata command. 

{pstd}
The csv, excel or text document created with the using subcommand can be used either directly or be linked to another master excel document. This way, the master document can be formatted and edited 
independantly while the excel documents generated by the command are updated. dltable table can also be used with listtex to create Latex tables.  

{pstd}
{cmd:dltable } accepts a qvalue option to account for multi-hypothesis testing. The pvalue used for the multi-hypothesis testing originates from the varlist regressions. qvalue accepts many different methods (see qvalue option). 
	
{title:Examples}

{phang}
Perform a simple ITT regression on several dependant varibles : {break}
{cmd:. dltable unemployment_rate health_score cognitive_score noncognitive_score, treat(treat_variable)}

{phang}
Save the results in a xlsx document :  {break}
{cmd:. dltable unemployment_rate health_score cognitive_score noncognitive_score using "myxlsx.xlsx" , treat(treat_variable) control(i.country)}

{phang}	
	Perform a simple LATE regression on several dependant varibles : {break}
 {cmd:. dltable unemployment_rate health_score cognitive_score noncognitive_score, treat(treat_variable) estimator(LATE) treated(treated_variable)}
 


{title:References}

{p 4 6 2}
Benjamini, Yoav and Krieger, Abba M and Yekutieli, Daniel (2006), Adaptive linear step-up procedures that control the false discovery rate, Biometrika, vol.93, n 3, pages 491--507,{p_end} 
{p 4 6 2}
Michael L. Anderson (2008), Multiple Inference and Gender Differences in the Effects of Early Intervention: A Reevaluation of the Abecedarian, Perry Preschool, and Early Training Projects},
Journal of the American Statistical Association, volume 103, number 484, pages 1481-1495. {p_end} 
{p 4 6 2}
Belloni, Alexandre, Victor Chernozhukov, and Christian Hansen. "Inference on treatment effects after selection among high-dimensional controls." Review of Economic Studies 81.2 (2014): 608-650. {p_end} 

 Adrien Bouguen, Santa Clara University