Link Individuals to daily dates.

[Heinrich Zitzewitz]

Hi fellows, I have a paneldataset of 56.000 rows and I am trying to implement a DiD regression including fixed effects.

My problem is that I have daily output data for 20 banks over a time period of 10 years. I am struggling with connecting the individual banks to their timeperiods.
The image displays the Bankid (no. 12) the DealAmount and the daily dates.

I have 20 banks and examine the daily output for 10 years as seen in the example.

I tried the following code, but honestly I do not really know how this works and where this leads me to.

egen id = seq()
replace id=id-1
gen id_temp=id/7
drop id_temp
egen id2=anymatch(id_temp), values(0(1)999)
gen id3 =id_temp if id2==1
replace id3=id3[_n-1] if missing(id3)
egen year=seq(), by (id3)
rename id3 persod
drop id_temp
drop id2
xtset persod year
xtreg DealAm TotalA i.year, fe cluster(persid)

Much appreaciate your help guys and girls.

[Heinrich Zitzewitz]

updated image. Sorry for that

[Ken Chui]

This question is probably impossible to address unless it's revised to be compliant with the FAQ (http://www.statalist.org/forums/help), particularly on using -dataex- to post the data in code form rather than in image form. It's very easy:

sysuse auto, clear

* To export 20 cases for variables mpg price and weight:
dataex price mpg weight, count(20)

And then post the results from the output, from CODE to /CODE

* Example generated by -dataex-. For more info, type help dataex
clear
input int(price mpg weight)
 4099 22 2930
 4749 17 3350
 3799 22 2640
 4816 20 3250
 7827 15 4080
 5788 18 3670
 4453 26 2230
 5189 20 3280
10372 16 3880
 4082 19 3400
11385 14 4330
14500 14 3900
15906 21 4290
 3299 29 2110
 5705 16 3690
 4504 22 3180
 5104 22 3220
 3667 24 2750
 3955 19 3430
 3984 30 2120
end

Do that twice, once for your bank data, once for your trading data. List only the variables that are involved in the merging. With those codes users here can create data sets and test their codes. It'd be easier for everyone to continue the dialogue that way.

Also, the codes posted in #1 do not make much immediate sense, it'd be great if you can add a bit of verbal description on what you'd like to achieve.

[Heinrich Zitzewitz]

Hejo Ken, thank you very much for the reply. Since I am new I much appreciate any instruction on how to behave correct on this forum.
I entered the code, as seen below. Hope this is more helpful.

What I did in the beginning is setting up a group variable for the individual banks (bankid). Then I included treatment variables for treatment- and controlgroup.
Eventually I wanted to implement fixed effects for the variable DealAm, because the dealamounts of a banks might be biased.

In order to implement fixed effects I tried to set each year from 2010-2019 to the individuall bankid.

For example Bankid Year
1 2010
2011
2012
....
2019
2 2010
2011
2012
....
2019


12 2010
2011
....
2019



My problem is that, besidedes multicollinearity, that I have daily outcome dates and between 1000-5000 rows per bankid.
What I need to do is link the bankid to individuall dates between 2010-2019. Not annualy but daily.

* Example generated by -dataex-. For more info, type help dataex
clear
input double DealAm int(Date Year1) float bankid
8.900e+08 18288 2010 12
8.900e+08 18288 2010 12
7.800e+09 18291 2010 12
7.800e+09 18291 2010 12
7.800e+09 18291 2010 12
2.500e+10 18305 2010 12
2.500e+10 18305 2010 12
2.500e+10 18305 2010 12
2.500e+10 18305 2010 12
2.500e+10 18305 2010 12
2.100e+09 18308 2010 12
2.100e+09 18308 2010 12
2.100e+09 18308 2010 12
3.500e+08 18311 2010 12
5.000e+09 18319 2010 12
5.000e+09 18319 2010 12
5.000e+09 18319 2010 12
3.018e+09 18326 2010 12
3.018e+09 18326 2010 12
3.018e+09 18326 2010 12
end
format %tdnn/dd/CCYY Date

[Ken Chui]

Thanks for the post. To be very honest I am still very lost. You mentioned that there is something to be linked, but I only saw that one file with daily data already.

If you meant that you wanted to create an empty daily series that runs from 2010 to 2019, for twelve times, one for each of the twelve banks. Then you can set up that series using this code:

clear
set obs 3652
gen date = mdy(12,31,2009) + _n
format date %td

expand 12

bysort date: gen bankid = _n
gsort bankid date

But again, I am not sure if that is what you want. Your question kept hinting that there is another data set that is not daily, but I couldn't see that.

Sorry, could be me being too tired.

[Jared Greathouse]

I don't understand the problem. If you can make the panel dataset, unless I'm missing something, isn't this just a 1:1 merge problem? Your unit and time ID's can just be merged together.