Carryforward/backward

[Marco Errico]

Dear all,

I have this panel data (I'm giving the example only about one company)

* Example generated by -dataex-. To install: ssc install dataex
clear
input long gvkey float(modate trn) byte(ffind48 tic_newind) float(event_time mean_trn treated treat_matching) int matching
29517 432   .08366858 42 0  .         . . .  .
29517 433    .1074952 42 0  .         . . .  .
29517 434   .04250479 42 0  .         . . .  .
29517 435   .10434986 42 0  .         . . .  .
29517 436   .11772685 42 0  .         . . .  .
29517 437   .05512161 42 0  .         . . .  .
29517 438   .05692236 42 0  .         . . .  .
29517 439  .072848454 42 0 -5 .09932967 1 1  .
29517 440   .07011225 42 0 -4 .09870233 1 1 13
29517 441   .05203461 42 0 -3 .12980506 1 1  .
29517 442  .072427504 42 0 -2 .12488867 1 1  .
29517 443   .06281572 42 0 -1  .1348747 1 1  .
29517 444   .07724509 13 1  0 .12799968 1 .  .
29517 445   .03720767 13 0  1 .12288648 1 .  .
29517 446  .023175865 13 0  2 .14432235 1 .  .
29517 447    .0207203 13 0  3  .1344712 1 .  .
29517 448   .04368569 13 0  4  .1244448 1 .  .
29517 449   .04155753 13 0  .         . . .  .
29517 450   .04167446 13 0  .         . . .  .
29517 451   .04698316 13 0 -5 .13748688 0 0  .
29517 452   .05114593 13 0 -4 .13370831 0 0  .
29517 453   .01690909 13 0 -3 .15079726 0 0  .
29517 454  .024613636 13 0 -2 .13852969 0 0  .
29517 455   .05386364 13 0 -1 .14454024 0 0  .
29517 456   .01256818 13 0  0  .1533158 0 .  .
29517 457  .026136363 13 0  1 .13952337 0 .  .
29517 458   .04420454 13 0  2  .1593396 0 .  .
29517 459   .03952273 13 0  3 .14585368 0 .  .
29517 460   .08820455 13 0  4 .14730963 0 .  .
29517 461   .07636364 13 0  .         . . .  .
29517 462   .14918615 13 0  .         . . .  .
29517 463   .04667914 13 0 -5 .09932967 1 1  .
29517 464  .033657625 13 0 -4 .09870233 1 1  .
29517 465   .04432392 13 0 -3 .12980506 1 1  .
29517 466   .08251356 13 0 -2 .12488867 1 1  .
29517 467   .04787732 13 0 -1  .1348747 1 1  .
29517 468   .04398241  2 1  0 .12799968 1 .  .
29517 469   .04447701  2 0  1 .12288648 1 .  .
29517 470    .1022898  2 0  2 .14432235 1 .  .
29517 471   .05145524  2 0  3  .1344712 1 .  .
29517 472  .010287388  2 0  4  .1244448 1 .  .
29517 473   .05271829  2 0  .         . . .  .
29517 474    .0877647  2 0  .         . . .  .
29517 475   .02890498  2 0 -5 .13748688 0 0  .
29517 476    .0880543  2 0 -4 .13370831 0 0  .
29517 477   .04041629  2 0 -3 .15079726 0 0  .
29517 478  .008470588  2 0 -2 .13852969 0 0  .
29517 479   .10133937  2 0 -1 .14454024 0 0  .
29517 480    .1401086  2 0  0  .1533158 0 .  .
29517 481    .1040181  2 0  1 .13952337 0 .  .
29517 482   .14177376  2 0  2  .1593396 0 .  .
29517 483  .063040726  2 0  3 .14585368 0 .  .
29517 484   .02276923  2 0  4 .14730963 0 .  .
29517 485  .029085973  2 0  .         . . .  .
29517 486  .014914027  2 0  .         . . .  .
29517 487  .026244344  2 0 -5 .13748688 0 0  .
29517 488   .03799095  2 0 -4 .13370831 0 0  .
29517 489   .01383908  2 0 -3 .15079726 0 0  .
29517 490   .02289655  2 0 -2 .13852969 0 0  .
29517 491   .06082759  2 0 -1 .14454024 0 0  .
29517 492  .013210728  2 0  0  .1533158 0 .  .
29517 493  .027417624  2 0  1 .13952337 0 .  .
29517 494  .024934866  2 0  2  .1593396 0 .  .
29517 495  .015632184  2 0  3 .14585368 0 .  .
29517 496  .019846743  2 0  4 .14730963 0 .  .
29517 497  .009655172  2 0  .         . . .  .
29517 498 .0012873564  2 0  .         . . .  .
29517 499 .0011340997  2 0 -5 .13748688 0 0  .
29517 500  .012291187  2 0 -4 .13370831 0 0  .
29517 501  .009230769  2 0 -3 .15079726 0 0  .
29517 502   .01276466  2 0 -2 .13852969 0 0  .
29517 503   .07029703  2 0 -1 .14454024 0 0  .
29517 504   .08819497  2 0  0  .1533158 0 .  .
29517 505    .1081188  2 0  1 .13952337 0 .  .
29517 506   .13154608  2 0  2  .1593396 0 .  .
29517 507    .0531297  2 0  3 .14585368 0 .  .
29517 508   .03775203  2 0  4 .14730963 0 .  .
29517 509  .033689436  2 0  .         . . .  .
29517 510   .11933265  2 0  .         . . .  .
29517 511   .16224205  2 0 -5 .13748688 0 0  .
29517 512   .10275135  2 0 -4 .13370831 0 0  .
29517 513   .23134495  2 0 -3 .15079726 0 0  .
29517 514   .21160975  2 0 -2 .13852969 0 0  .
29517 515    .2663972  2 0 -1 .14454024 0 0  .
29517 516    .3348346  2 0  0  .1533158 0 .  .
29517 517   .13191105  2 0  1 .13952337 0 .  .
29517 518   .20818377  2 0  2  .1593396 0 .  .
29517 519    .2983787  2 0  3 .14585368 0 .  .
29517 520    .6756593  2 0  4 .14730963 0 .  .
29517 521    .7332361  2 0  .         . . .  .
29517 522    .9892002  2 0  .         . . .  .
29517 523    .5725882  2 0 -5 .13748688 0 0  .
29517 524    .3594635  2 0 -4 .13370831 0 0  .
29517 525    .6906328  2 0 -3 .15079726 0 0  .
29517 526    .8219657  2 0 -2 .13852969 0 0  .
29517 527   .33211195  2 0 -1 .14454024 0 0  .
29517 528    .5735751  2 0  0  .1533158 0 .  .
29517 529    .3689737  2 0  1 .13952337 0 .  .
29517 530    .3584927  2 0  2  .1593396 0 .  .
29517 531    .2158612  2 0  3 .14585368 0 .  .
end
format %tm modate

What I need to do is to carryforward the values on variable "matching" as long as the variable treated is either 1 or 0 (no missing). As you can see from the example, I also have to carrybackward the same value (again ,as long as the variable treated is either 1 or 0)
So i know, that carryforward command is on ssc and sometimes it might be a misleading option. Neverthless, I don't see other solutions in coding it.

So far, what I've done is the following

bys gvkey (modate): carryforward matching if treated==0 | treated==1, gen(match1)

But, what about to carrybackward? I tried with the option "back" without valuable results.

Do you have other options? (I'm happy also without using carryforward command )

Many thanks in advance

[Ali Atia]

The carryforward help file provides instructions on how to do this. You need to generate a negative date variable:

gen int negmodate = neg(modate)

Then:

bys gvkey (negmodate): carryforward matching if treated==0 | treated==1, gen(match2)

And you can combine the two variables using

replace match1 = match2 if match1 == .
drop match2

[Nick Cox]

I don't see a complete specification of rules here. This may be a start. I

clonevar wanted = matching 
bysort gvkey (modate) : replace wanted = wanted[_n-1] if missing(wanted) & inlist(treated, 0, 1)    
gsort gvkey -modate 
by gvkey : replace wanted = wanted[_n-1] if missing(wanted) & inlist(treated, 0, 1)
sort gvkey modate

[Marco Errico]

The carryforward help file provides instructions on how to do this. You need to generate a negative date variable:

gen int negmodate = neg(modate)

Then:

bys gvkey (negmodate): carryforward matching if treated==0 | treated==1, gen(match2)

And you can combine the two variables using

replace match1 = match2 if match1 == .
drop match2

Thanks Ali Atia . Nevertheless, when I try with:

	
	
gen int negmodate = neg(modate)

"unknown function neg()
r(133)"

I also tried with egen, but same thing. Maybe is it a specific package? I'm using STATA 15.1

Thanks

[Marco Errico]

Thanks Nick Cox

Your codes really work properly without using carryfroward. Apologizes if I wasn't clear in the explanation: I only wanted to extend the variable matching as long as there is a value (1-0) in the variable treated.
I tried many days in many ways. I didn't know that the option "inlist" can be really helpful in this case.

Thanks again for your time.

Marco

[Nick Cox]

Thanks for the thanks. To avoid confusion, note that inlist() is a function, not an option. In this case

(treated == 0 | treatment == 1)

would be fine too.

[Marco Errico]

Thanks! I'm already studying it.

Sorry for a quick question.
Regarding your codes, when you suggest

bysort gvkey (modate) : replace wanted = wanted[_n-1] if missing(wanted) & inlist(treated, 0, 1)

would lead to the same outcome using

bysort gvkey (modate) : replace wanted = wanted[_n-1] if missing(wanted) & treated==0 | treated==1

?

EDIT: didn't know you already answered to this question Apologize

[Nick Cox]

 
 missing(wanted) & (treated==0 | treated==1)

spells out what you want more explicitly. I can never remember the precedence rules, so tend just to parenthesise aggressively.

[Ali Atia]

Sorry about that Marco Errico - the proper command is

-modate

, not

neg(modate)

. Either way, it looks like you were able to figure it out with Nick's help.

[Marco Errico]

Yes thanks Ali Atia and Nick Cox .

Finally solved, also without using carryforward command that sometimes lead to puzzling results