command Import don't work in version 11.

Investigate if we can substitute for insheet or if we need to set ietoolkit to require version 13

simply encode it and do not allow any of option that is making references to the codes.

Options that will not be allowed if string variable

• control(groupcode)
• order(groupcodelist)
• grpcodes
• grplabels(codetitles)

Shouldn't the value for obs excluded by maxdiff() have a missing value in _matchDiff

When specifying robust in the vce() option an error is thrown saying "vce(robust) is not allowed". vce() works wioth cluster var.

I'm doing balance tests on a single variable for different values of a categorical variable, which means using different if statements. I couldn't find a quick way to do that and have them all in a single table. Is there a way to do that?

If not, maybe it could be interesting to add something like an over(varname) option? Or simply add append as an option when saving the file?

The labels (1)-(2), (1)-(3) etc. in the test column is not showing when using the control() option. This only happens when exporting to TeX, not when export to Excel or browsing

File paths in save() and savetex() is not working in Mac. It save the full name of the path as file name in the current directory.

Please find iegitaddtxt.ado in the develop-iegitaddtxt branch. See iegitaddtxt.sthlp in the same branch for the helpfile.

@mrimal please test this command for bugs and errors

@luizaandrade or anyone else, feel free to test it you too if you have some time to spare!

Report any bugs by writing to the branch (or create a pull request) or by commenting in this thread and I will fix.

Put the master folders in the same folder as round, but add round and master to the names

The idvar option does not allow for a varlist even though it is specified in the help file.

Strange error message with option save() is not used

== 0 is not a command

Thank you for notifying me Wenqing!

I was just asked to rerun balance checks weighting observations and realized iebaltab doesn't allow that option.

If clusters are used, show the number or clusters per group.

Is there a way to manually name the file of duplicate reports? Suppose we use “ieduplicates” command for many different dataset, then we need to save each report separately.

In addition to be a great way of documenting the intended use of the folder structure it would also be a sole the problem of github not syncing empty folders

So that one does not have to run the project master to run the for example the baseline master dofile

In projects with many rounds there are many places where all users have to specify their root folder path.

By handling root folders in the global_setup.do file and calling that file in each master do file, it will limit the places this have to be specified to one, and it won't conflict with the preference that lead to issue #80

The command does not automatically wrap text when varlabels are especially long and exporting to table to LaTeX. This can be manually solved in LaTeX but there is no option which can do it through iebaltab.

Thanks.

It would be great if there could be variable labels for the following variables that are generated once you run the iematch command:
_matchResult
_matchID
-matchDiff
-matchCount

When we are testing which file extension that is used in the save() option we are looking for the dot and compare the rest of the string to the file extensions allowed. However, it is possible to have a . in a folder name or anything meaning that there will be a . in the middle of the file path in save(). Then we are not comparing the file extension but a large part of the file path to the allowed file extensions and we will always get an error.

Does this apply to savetex() as well?

This problem should also be the case in commands with similar code, for example iegraph.

I wanted to have a specific value labeled on the second y-axis on the graph and therefore tried to generate the second y-axis using yaxis(1 2). But an error happened saying "option yaxis() not allowed". The same situation happened to barwidth.

Thanks,
Wenqing

Sub folders are created with the parent folder in the name.
Subsequent usages of the command therefore do not find the folders to create folders in

If rows in the duplicates report has been deleted, for any reason, then the report might not be read properly. Test the input and give an informative answer if this is the case.

when a label as a $ inside a word, like in Fanta$tic. Then the escape character does not work. When numbers follows like $2000 then it works.

I have only tested this in the table notes. Not in any other label

You can add this to README. Documenting the developing environment is a good practice in general.

In order for the graph to be valid each dummy used in varlist should be a dummy per treatment arm. And there can be anywhere from one to many treatment arms. Since no observation can be in two treatment arms, then there should be no overlap. And some observations, i.e. the control, should have zeros in all dummies. We should test that this is the case, currently we are only testing that they are dummies.

The only thing that makes this difficult is that we also want this command to be working for diff-in-diff regressions. In a diff-in-diff there are exactly three dummies. One of those dummies should be the product of the other two. And then there must be observations in all groups, [0,0], [0,1], [1,0] and [1,1].

And we need to have an error message that explains why we are doing this test.

We should do this, but this does not have to be done in the update we are planning to send in end of July.

This is on our list and will hopefully be included in next version of ietoolkit. If you have resources for your favorite standardized format for a balance table, then please post them below. There are many different schools on how tables should be formatted and deciding for one of them will be what takes us the most time to add an option for latex to iebaltab.

I was trying to create a graph with a long title but the title cannot be shown in full length. So I wrote the code below:

iegraph treat, ///
title ("how easy or difficult is it to find out how the LGA is spending its budget" ///
, size(small) position(11))

But the title doesn't change at all. I'm wondering how can I adjust the size and style of the title?

Thanks,
Wenqing

In an earlier version of ieboilstart the command created a dofile used to set the settings. This is no longer the case but there this dofiel is still mentioned in the help file.

If a variable name is used as row title and that variable name includes and underscore, for example d_harverst, then the "h" will be subscripted to the "D".

They way to get around that now is variable a variable label and use option rowvarlabel, or to give it another label using the rwolabel() option. But if there is an easy way to include an escape character before the underscore, then that would be great.

It seems that in “ieduplicates” helpfile, the description of “folder” option is missing under “Syntax” category (There is description of “folder” option under “Options” category below, but not in “Syntax” category).

Instead of exporting a file with the document settings etc. commented out we should have an option that export only the table without those comments so that it can be used to import in our DIME LaTeX template.

Thanks for doing this @luizaandrade !

Add option to add readme placeholder to every subfolder. Syntax would be:

(line 7) syntax , folder(string) [all]
(line 25) if (`"`flist'`dlist'`olist'"' == "") | ("`all'" == "all") { 

In a regression on a variable where the mean (for control) and the mean+beta (for treatment arms) are negative yzero() creates an error. yzero() looks for the maximum value to determine the length of the axis but that is only correct for positive bars. Negative bars should use the minimum value.

This is most likely easily solved with an if-statement based on a test if the values for the bars are positive or negative. The only special case is when some bars are positive and some are negative. But in that case yzero() does not make sense as the 0 on the y-axis is included anyways.

So the solution to this issue is that if yzero() is used, then we should test that the value for all bars (and confidence intervals if used) are all positive or negative. If they are not then throw an error saying that the options cannot be used. If error is not thrown, then test if they are all positive and use maximum, otherwise use minimum.

Some str vars (but strangely enough not all) in the if-statement to iedropone cases a type miss-match error.

If a variable name using multiple underscores then only one is escaped

The default setting is that the N is reported for each group for each variable. There is an option that puts the N at a row below all variables with just one N per group. This option is only possible if the N is the same within each group across all variables.

Currently the help file calls this onerown but the code expect the name to be onenrow. What is the best name for this option? onerown or onenrow? Or is nonerow better? Or something completely different?

@luizaandrade , @eatorange , @morevba @mrimal do either of you have an opinion of what is the best name for this option?

Instead of throwing an error or requiring nodaily to be turned on, just use the command mkdir to create it.

nodaily can be kept as an option, but instead of that option disabling the requirement it should instead be that it disabling creating a new folder if it does not exist

In large and complex projects we might have 5 data sources in one round, and maybe have 5 rounds. That creates too many folders. One solution would be to be able to create a round called, f.ex., Baseline which is only a folder and a master dofile. Then create folders inside that folder. The key word round in the command would then be reserved for creating this type of folder, and a new key word source would be used to create what is now called round. For a source you have the possibility to add the key word in that allows you to create that source in that round that is specified. See an example how it would be specified below.

*Create a new project
iefolder new project                        , projectfolder("$projectABC")
	
*Add a unit of observation
iefolder new unitofobs household 	    , projectfolder("$projectABC")
	
*Create a new round called baseline
iefolder new round Baseline	            , projectfolder("$projectABC") 

*Create two new sources in Basline folder
iefolder new source farmer_BL in Baseline   , projectfolder("$projectABC") abbreviation("farmerBL")
iefolder new source hhh_BL in Baseline	    , projectfolder("$projectABC") abbreviation("hhhBL")

*Create a new sources that is not Baseline
iefolder new source monitoring              , projectfolder("$projectABC") abbreviation("monit")

This is not a project for right now, but lets keep this as a place where we can discuss how it should work, and how to implement it. So feel free to brainstorm features, challenges and solutions.

This idea came up in a discussion with @gbedoyaWB

Please proofread helpfile iegitaddtxt.sthlp in branch develop-iegitaddtxt.

I will test the command a bit more and then afterwards ask you to also test it.

Thanks!

In "ieboilstart"'s help file, "custom" option refers to "example 4" as below,

custom(string) allows the user to add one or multiple custom lines of code. Each line of code should be seperated with a "@". See example 4 below for more details.

However, there's no example 4 in the help file. I believe it should be example 2, which shows an example of "custom" option as below.

ieboilstart, versionnumber(12.1) custom(ssc install estout @ ssc install winsor)
`r(version)'

When using % in variable labels, TeX reads it as a comment. Add escape character.

A test feature was incorrectly left in the syntax for ieboilstart. This bug was fatal but is now fixed for version 2.1 of ietoolkit. The update has been submitted to the server, and this issue will be closed as soon as the update is reflected on the server.

It is easy to implemented. The solution is to not hard code the value 1.96. (1.96 is only accurate in large samples anyways)

This section:

scalar  conf_int_min_`var'   =	coeff_`var'-(1.96*coeff_se_`var') + ctl_mean
scalar  conf_int_max_`var'   =	coeff_`var'+(1.96*coeff_se_`var') + ctl_mean

Change to:

*from the regression:
local df = `e(df_r)'

*From option or set default
if "confbarval" = "" {
	local confbarval = .95
}

*Go to one tail value
local confbarval_1tail = ( `confbarval' + (1-`confbarval' ) / 2)

*Calculate t-stats
local tstats = invt(`df' , `confbarval_1tail'  )

*Use it in the old code
scalar  conf_int_min_`var'   =	coeff_`var'-(`tstats' * coeff_se_`var') + ctl_mean
scalar  conf_int_max_`var'   =	coeff_`var'+(`tstats' * coeff_se_`var') + ctl_mean

What requires a little bit more thinking is how to document this in the output.

Reminder to myself to update the iegitaddmd help file before publishing

In order to make the round master do-files independent from the project master do-file then it also needs to have the unit global section. This issue is related to issue #80

Not great to have to define this at multiple location, but at least it is only at a few places.

Perhaps this can be solved by storing these in a separate file that is run in the different locations.

Allow appending tables exported from iebaltab the way, for example, estout does it.

Is that something we can implement? Working with and modifying different file formats can be tricky, but if someone sees a straightforward way to implement this, then I would be happy to include it.

I guess it would only apply to tables exported to Excel. Or could this make sense in LaTeX too?

Based on feedback from @aidancoville.