Event Based Model with not estimable margin problem

[amera amery]

Hello All,
I am building event based model to check the effect of Covid-19 policies in US. I built it using fixed effect regression. the model is studying the impact of mask policy in counties that applied business close or not. after building the regression i got results and coefficient as follows:







I see the coefficient for last two days are omitted. when I tried to get margin using margin command I got (not estimable) message for all coefficients. I read a lot on your website and see one solution by using noestimcheck margin. I got results for margin after using this command. My question is if I can adopt these results of margin on my research. is it accurate results that I can use it? I appreciate your help as I need help to complete my research and submit my results after being sure about it.


My code:
g busniss_policy=business_delta_day


replace busniss_policy=1 if busniss_policy>=0

replace busniss_policy=0 if busniss_policy<0


* to rprint coeff significant statistics->postestimator->managepst results->table by adding business_delta_day>0 means these counties applied business close policy



reghdfe cases_normed i.shifted_ttt11#busniss_policy cases_normed_delay_14days cases_normed_delay_14days_growth residential workplaces transit_stations parks grocery_and_pharmacy retail_and_recreation grocery_and_pharmacy_sq retail_and_recreation_sq parks_sq transit_stations_sq workplaces_sq if shifted_ttt11<=368 & shifted_ttt11>=331&mask!="2077-12-31", absorb(fips time) vce(cluster fips)

margins shifted_ttt11#busniss_policy,noestimcheck

[Clyde Schechter]

In general, when Stata tells you things are not estimable, you will get unreliable results by forcing estimation with the -noestimcheck- option. There are exceptions, but your situation is not one of them.

That said, I think you have not set up your regression properly to use -margins- anyway. I'm guessing from the names of your variables that things like parks_sq represent the square of parks (i.e. parks^2). If I have that right, you cannot use -margins- with regression, and you were, in a way, lucky that something turned out inestimable so you didn't just blunder on using meaningless results. If you want to use -margins-, you cannot use "homebrew" quadratic variables in your regression. Rather, you must use factor variable notation to create them. Instead of having var and var_sq, you must use c.var##c.var (N.B. ##, not #). Stata will then internally generate the linear and quadratic terms and include them in the regression. And, importantly, that will assure that -margins- knows that the quadratic term is a quadratic term and will handle it correctly in the calculations. (When you use var and var_sq, Stata thinks that these are two unrelated variables and therefore mishandles the calculations involving them.)

There may be other problems not apparent from what you showed. The screenshot of your results is unreadable on my setup. Unreadability of screenshots is very common, which is why the Forum FAQ advises against using them. The effective way to show Stata results is to copy them from the Results window or your log file into the Forum editor and surround them by code delimiters. Read FAQ #12 if you are not familiar with code delimiters. (I imagine you aren't since you didn't use them for your code either.) Also, the code you show is evidently not your real code, because some of the variable names that appear in it are far longer than the 32 character limit on variable names in Stata. So I don't want to go too deep into trying to analyze your code, as I don't know in what ways it differs from the real thing.

[amera amery]

I have removed the variables you mentioned and just use google mobility data. to show that the problem is not with variables. I tried to build the intersection fixed effect regression model (day#busniss_policy) using xtreg where days is number for days after applying mask policy and busniss_policy is (0,1) indicates if the county apply busniss_policy or not. The results of regression as below where all coefficients are estimated except for last day (day 368)

xtset fips time

replace  delta_mask= 0 if mask=="2077-12-31"

summ delta_mask
g days= delta_mask - r(min)



summ days  if delta_mask == -7


* used to find the index of -7 day after shifting and 30 days after policy 
local true_neg1 = r(mean)

summ days if delta_mask== 30


g busniss_policy=business_delta_day


replace busniss_policy=1 if busniss_policy>=0

replace busniss_policy=0 if busniss_policy<0


reghdfe cases_normed i.days#busniss_policy cases_normed_delay_14days cases_normed_delay_14days_growth residential workplaces transit_stations parks grocery_and_pharmacy retail_and_recreation  if days<=368 & days>=331&mask!="2077-12-31", absorb(fips time) vce(cluster fips) asis

margins  days#busniss_policy, post

Results:
(dropped 166 singleton observations)
(MWFE estimator converged in 24 iterations)
note: 368.days#0b.busniss_policy omitted because of collinearity
note: 368.days#1.busniss_policy omitted because of collinearity

HDFE Linear regression Number of obs = 3,590
Absorbing 2 HDFE groups F( 81, 587) = 3.24
Statistics robust to heteroskedasticity Prob > F = 0.0000
R-squared = 0.6365
Adj R-squared = 0.5250
Within R-sq. = 0.0593
Number of clusters (fips) = 588 Root MSE = 0.4205

(Std. err. adjusted for 588 clusters in fips)
--------------------------------------------------------------------------------------------------
| Robust
cases_normed | Coefficient std. err. t P>|t| [95% conf. interval]
---------------------------------+----------------------------------------------------------------
days#busniss_policy |
331 0 | 0 (base)
331 1 | -.3785024 .1439046 -2.63 0.009 -.6611329 -.0958719
332 0 | -.0286351 .0810405 -0.35 0.724 -.1877999 .1305296
332 1 | -.4687431 .1248702 -3.75 0.000 -.7139898 -.2234963
333 0 | -.0529339 .0696404 -0.76 0.447 -.1897086 .0838409
333 1 | -.4350848 .195188 -2.23 0.026 -.8184367 -.0517328
334 0 | -.0123204 .0716729 -0.17 0.864 -.153087 .1284462
334 1 | -.2812313 .1389139 -2.02 0.043 -.5540602 -.0084025
335 0 | -.0948619 .0578722 -1.64 0.102 -.2085236 .0187998
335 1 | -.4375989 .1179329 -3.71 0.000 -.6692206 -.2059771
336 0 | .0984214 .0709704 1.39 0.166 -.0409654 .2378082
336 1 | -.3576105 .136115 -2.63 0.009 -.6249422 -.0902787
337 0 | -.0111905 .0620542 -0.18 0.857 -.1330657 .1106847
337 1 | -.3619055 .124545 -2.91 0.004 -.6065136 -.1172974
338 0 | .0315402 .0512197 0.62 0.538 -.069056 .1321364
338 1 | -.2556062 .1190892 -2.15 0.032 -.4894989 -.0217135
339 0 | .0159766 .0648646 0.25 0.806 -.1114183 .1433714
339 1 | -.2983611 .1595379 -1.87 0.062 -.6116958 .0149736
340 0 | -.0212477 .0693547 -0.31 0.759 -.1574612 .1149659
340 1 | -.3429085 .1118587 -3.07 0.002 -.5626004 -.1232166
341 0 | -.0154597 .058877 -0.26 0.793 -.1310949 .1001755
341 1 | .0518958 .1552399 0.33 0.738 -.2529974 .3567891
342 0 | .0872814 .0678902 1.29 0.199 -.0460559 .2206187
342 1 | -.2198898 .1176075 -1.87 0.062 -.4508725 .0110928
343 0 | .0559241 .0566958 0.99 0.324 -.0554272 .1672755
343 1 | -.0154187 .2057291 -0.07 0.940 -.4194735 .3886361
344 0 | .098108 .1115979 0.88 0.380 -.1210719 .3172878
344 1 | -.209594 .103771 -2.02 0.044 -.4134017 -.0057863
345 0 | .0253598 .0568927 0.45 0.656 -.0863783 .137098
345 1 | -.2047456 .1220154 -1.68 0.094 -.4443855 .0348943
346 0 | .0047448 .0730229 0.06 0.948 -.1386731 .1481626
346 1 | -.2253762 .1352034 -1.67 0.096 -.4909175 .0401652
347 0 | .0901131 .060661 1.49 0.138 -.0290259 .2092522
347 1 | -.2708912 .1155353 -2.34 0.019 -.4978042 -.0439783
348 0 | .0463957 .0603709 0.77 0.442 -.0721735 .164965
348 1 | -.0800682 .1277212 -0.63 0.531 -.3309144 .170778
349 0 | .1748163 .059026 2.96 0.003 .0588885 .2907441
349 1 | -.1932883 .1363676 -1.42 0.157 -.4611162 .0745395
350 0 | .1276891 .0639394 2.00 0.046 .0021112 .2532669
350 1 | -.2461021 .1353546 -1.82 0.070 -.5119403 .0197362
351 0 | -.0484437 .0612692 -0.79 0.429 -.1687773 .0718898
351 1 | -.1128159 .1252921 -0.90 0.368 -.3588913 .1332594
352 0 | .1978287 .0788838 2.51 0.012 .0428999 .3527574
352 1 | -.3310455 .1166044 -2.84 0.005 -.5600581 -.1020328
353 0 | .0318333 .062043 0.51 0.608 -.09002 .1536865
353 1 | -.1871617 .1164498 -1.61 0.109 -.4158708 .0415473
354 0 | .1856217 .0779045 2.38 0.018 .0326162 .3386272
354 1 | -.1416514 .1101104 -1.29 0.199 -.3579096 .0746069
355 0 | .0962708 .0568877 1.69 0.091 -.0154574 .2079989
355 1 | -.2366667 .121115 -1.95 0.051 -.4745382 .0012048
356 0 | .2624909 .1101126 2.38 0.017 .0462283 .4787534
356 1 | -.2017364 .145319 -1.39 0.166 -.4871449 .0836721
357 0 | .1301716 .0568979 2.29 0.023 .0184234 .2419198
357 1 | -.1322139 .1097681 -1.20 0.229 -.3478 .0833721
358 0 | .1050809 .0531327 1.98 0.048 .0007275 .2094343
358 1 | -.1418137 .1130372 -1.25 0.210 -.3638203 .0801929
359 0 | .165354 .0600988 2.75 0.006 .0473191 .2833888
359 1 | -.1778536 .1022634 -1.74 0.083 -.3787002 .022993
360 0 | .0910074 .061387 1.48 0.139 -.0295575 .2115722
360 1 | -.1071111 .1158942 -0.92 0.356 -.3347288 .1205066
361 0 | .207608 .0697663 2.98 0.003 .0705861 .3446299
361 1 | -.2146253 .1224156 -1.75 0.080 -.4550511 .0258006
362 0 | -.0021286 .073323 -0.03 0.977 -.1461361 .1418788
362 1 | -.3290501 .1366439 -2.41 0.016 -.5974205 -.0606797
363 0 | .0857311 .0621154 1.38 0.168 -.0362644 .2077265
363 1 | -.140684 .0999765 -1.41 0.160 -.3370392 .0556711
364 0 | .0868201 .0812561 1.07 0.286 -.0727681 .2464083
364 1 | -.0915501 .0974802 -0.94 0.348 -.2830025 .0999022
365 0 | .0078131 .0750609 0.10 0.917 -.1396075 .1552337
365 1 | .0100225 .1258028 0.08 0.937 -.2370559 .2571009
366 0 | .052421 .0706834 0.74 0.459 -.0864023 .1912442
366 1 | -.0800325 .1515284 -0.53 0.598 -.3776364 .2175715
367 0 | .0259653 .0645145 0.40 0.687 -.1007421 .1526726
367 1 | -.0761031 .0832521 -0.91 0.361 -.2396115 .0874052
368 0 | 0 (omitted)
368 1 | 0 (omitted)
|
cases_normed_delay_14days | .1332996 .0566353 2.35 0.019 .0220671 .244532
cases_normed_delay_14days_growth | -.002791 .0005906 -4.73 0.000 -.0039509 -.0016312
residential | -.7144175 .5993666 -1.19 0.234 -1.891582 .4627466
workplaces | -.3887375 .2089706 -1.86 0.063 -.7991586 .0216836
transit_stations | .0438485 .0416414 1.05 0.293 -.0379358 .1256328
parks | -.0080359 .0261777 -0.31 0.759 -.0594491 .0433774
grocery_and_pharmacy | .1834313 .1586384 1.16 0.248 -.1281366 .4949993
retail_and_recreation | -.0696244 .1034822 -0.67 0.501 -.2728649 .1336161
_cons | .778438 .8210108 0.95 0.343 -.8340383 2.390914

After trying margin command

margins  days#busniss_policy, post

The results as follows:


-------------------------------------------------------------------------------------
| Delta-method
| Margin std. err. z P>|z| [95% conf. interval]
--------------------+----------------------------------------------------------------
days#busniss_policy |
331 0 | . (not estimable)
331 1 | . (not estimable)
332 0 | . (not estimable)
332 1 | . (not estimable)
333 0 | . (not estimable)
333 1 | . (not estimable)
334 0 | . (not estimable)
334 1 | . (not estimable)
335 0 | . (not estimable)
335 1 | . (not estimable)
336 0 | . (not estimable)
336 1 | . (not estimable)
337 0 | . (not estimable)
337 1 | . (not estimable)
338 0 | . (not estimable)
338 1 | . (not estimable)
339 0 | . (not estimable)
339 1 | . (not estimable)
340 0 | . (not estimable)
340 1 | . (not estimable)
341 0 | . (not estimable)
341 1 | . (not estimable)
342 0 | . (not estimable)
342 1 | . (not estimable)
343 0 | . (not estimable)
343 1 | . (not estimable)
344 0 | . (not estimable)
344 1 | . (not estimable)
345 0 | . (not estimable)
345 1 | . (not estimable)
346 0 | . (not estimable)
346 1 | . (not estimable)
347 0 | . (not estimable)
347 1 | . (not estimable)
348 0 | . (not estimable)
348 1 | . (not estimable)
349 0 | . (not estimable)
349 1 | . (not estimable)
350 0 | . (not estimable)
350 1 | . (not estimable)
351 0 | . (not estimable)
351 1 | . (not estimable)
352 0 | . (not estimable)
352 1 | . (not estimable)
353 0 | . (not estimable)
353 1 | . (not estimable)
354 0 | . (not estimable)
354 1 | . (not estimable)
355 0 | . (not estimable)
355 1 | . (not estimable)
356 0 | . (not estimable)
356 1 | . (not estimable)
357 0 | . (not estimable)
357 1 | . (not estimable)
358 0 | . (not estimable)
358 1 | . (not estimable)
359 0 | . (not estimable)
359 1 | . (not estimable)
360 0 | . (not estimable)
360 1 | . (not estimable)
361 0 | . (not estimable)
361 1 | . (not estimable)
362 0 | . (not estimable)
362 1 | . (not estimable)
363 0 | . (not estimable)
363 1 | . (not estimable)
364 0 | . (not estimable)
364 1 | . (not estimable)
365 0 | . (not estimable)
365 1 | . (not estimable)
366 0 | . (not estimable)
366 1 | . (not estimable)
367 0 | . (not estimable)
367 1 | . (not estimable)
368 0 | . (not estimable)
368 1 | . (not estimable)


Another question: How Can I plot the coefficient of results foe each group separately in same graph ( 0,1) groups of coefficients?

[Clyde Schechter]

Your problem is arising because of some colinearity. Notice that Stata warned you "note: 368.days#0b.busniss_policy omitted because of collinearity
note: 368.days#1.busniss_policy omitted because of collinearity." Since the pattern we see is the omission of the two terms involving days == 368, which is the largest value of the date variable, the likely cause is that there is some other variable in your model that is constant over some groups of dates. This is the usual cause of this kind of colinearity. Based on the names of your variables, I would guess that the cases_normed_delay_14_days or cases_normed_delay_14days_growth variable (or perhaps both) is the root of the problem here. That sounds like something that might be constant within values of variable days, or groups of such values. But, perhaps you can identify the offending variable easily because you are familiar with the data. You need to remove that variable from the model. Then you will get estimates for all of the days, and the -margins- problem will go away once that is resolved. It is not possible for -margins- to identify marginal effects for variables that participate in a colinearity.

How Can I plot the coefficient of results foe each group separately in same graph ( 0,1) groups of coefficients?

I don't understand the question. Which groups are you referring to, and what exactly do you want to plot: what goes on the horizontal axis, and what goes on the vertical axis?

[amera amery]

Thank you very much for your reply as I spent days trying to solve this problem. after removing some variables that may cause this problem i am still getting omitted coefficient for the last day as follows:
(dropped 166 singleton observations)
(MWFE estimator converged in 24 iterations)
note: 368.days#0b.busniss_policy omitted because of collinearity
note: 368.days#1.busniss_policy omitted because of collinearity

HDFE Linear regression Number of obs = 3,590
Absorbing 2 HDFE groups F( 79, 587) = 2.18
Statistics robust to heteroskedasticity Prob > F = 0.0000
R-squared = 0.6279
Adj R-squared = 0.5140
Within R-sq. = 0.0370
Number of clusters (fips) = 588 Root MSE = 0.4253

(Std. err. adjusted for 588 clusters in fips)
-------------------------------------------------------------------------------
| Robust
cases_normed | Coefficient std. err. t P>|t| [95% conf. interval]
--------------+----------------------------------------------------------------
days#|
busniss_pol~y |
331 1 | -.4099208 .1463507 -2.80 0.005 -.6973556 -.1224861
332 0 | -.0280733 .0837722 -0.34 0.738 -.192603 .1364564
332 1 | -.4984619 .1224785 -4.07 0.000 -.7390113 -.2579125
333 0 | -.0616372 .0700387 -0.88 0.379 -.1991941 .0759198
333 1 | -.5363874 .185835 -2.89 0.004 -.9013698 -.1714049
334 0 | -.0225765 .0739121 -0.31 0.760 -.1677409 .1225878
334 1 | -.3172599 .1398723 -2.27 0.024 -.5919711 -.0425487
335 0 | -.0933705 .0597221 -1.56 0.118 -.2106655 .0239244
335 1 | -.5004138 .1193204 -4.19 0.000 -.7347608 -.2660669
336 0 | .0967768 .0701418 1.38 0.168 -.0409828 .2345363
336 1 | -.4009202 .1458857 -2.75 0.006 -.6874417 -.1143988
337 0 | -.0125092 .0618269 -0.20 0.840 -.1339381 .1089197
337 1 | -.4015682 .1228618 -3.27 0.001 -.6428705 -.160266
338 0 | .0315365 .0510951 0.62 0.537 -.0688151 .1318881
338 1 | -.2958461 .1156456 -2.56 0.011 -.5229756 -.0687165
339 0 | .042306 .0674217 0.63 0.531 -.0901112 .1747232
339 1 | -.3720456 .1555316 -2.39 0.017 -.6775117 -.0665794
340 0 | -.0115566 .0715844 -0.16 0.872 -.1521494 .1290362
340 1 | -.3557411 .1119916 -3.18 0.002 -.5756941 -.135788
341 0 | -.0185854 .0583593 -0.32 0.750 -.1332039 .096033
341 1 | .0148648 .1622708 0.09 0.927 -.3038373 .333567
342 0 | .1023682 .0698598 1.47 0.143 -.0348373 .2395738
342 1 | -.2631451 .1163144 -2.26 0.024 -.4915882 -.0347021
343 0 | .0740682 .0575279 1.29 0.198 -.0389175 .1870538
343 1 | -.045251 .2053903 -0.22 0.826 -.4486403 .3581383
344 0 | .1176535 .1101376 1.07 0.286 -.0986582 .3339652
344 1 | -.2291802 .1020424 -2.25 0.025 -.4295929 -.0287674
345 0 | .0408045 .0563465 0.72 0.469 -.0698607 .1514698
345 1 | -.2384194 .1255983 -1.90 0.058 -.4850961 .0082573
346 0 | .0222127 .0756599 0.29 0.769 -.1263844 .1708097
346 1 | -.2546369 .1326608 -1.92 0.055 -.5151845 .0059107
347 0 | .1153542 .0639572 1.80 0.072 -.0102585 .240967
347 1 | -.3069842 .1106948 -2.77 0.006 -.5243902 -.0895781
348 0 | .0553402 .0593864 0.93 0.352 -.0612955 .1719759
348 1 | -.0976982 .1296248 -0.75 0.451 -.352283 .1568866
349 0 | .1930237 .0584035 3.30 0.001 .0783184 .307729
349 1 | -.2138118 .1374718 -1.56 0.120 -.4838084 .0561847
350 0 | .1445166 .0630535 2.29 0.022 .0206787 .2683546
350 1 | -.2743005 .128843 -2.13 0.034 -.52735 -.021251
351 0 | -.0289425 .0635516 -0.46 0.649 -.1537586 .0958737
351 1 | -.1453901 .1220631 -1.19 0.234 -.3851236 .0943435
352 0 | .214769 .0776344 2.77 0.006 .062294 .367244
352 1 | -.3472789 .1164588 -2.98 0.003 -.5760056 -.1185521
353 0 | .0450266 .0625553 0.72 0.472 -.0778328 .167886
353 1 | -.1798121 .1141795 -1.57 0.116 -.4040622 .0444379
354 0 | .2014349 .0754679 2.67 0.008 .053215 .3496548
354 1 | -.1796731 .1079783 -1.66 0.097 -.391744 .0323979
355 0 | .1087534 .054784 1.99 0.048 .0011569 .2163499
355 1 | -.2302023 .1196702 -1.92 0.055 -.4652361 .0048315
356 0 | .2824935 .107707 2.62 0.009 .0709554 .4940316
356 1 | -.2231089 .1538338 -1.45 0.148 -.5252405 .0790227
357 0 | .1547178 .0556484 2.78 0.006 .0454236 .2640121
357 1 | -.0782992 .1081413 -0.72 0.469 -.2906902 .1340919
358 0 | .1349423 .0531275 2.54 0.011 .0305992 .2392854
358 1 | -.1575049 .1127381 -1.40 0.163 -.3789242 .0639143
359 0 | .1751728 .0602025 2.91 0.004 .0569343 .2934113
359 1 | -.1489528 .0987957 -1.51 0.132 -.3429889 .0450833
360 0 | .1098763 .0616988 1.78 0.075 -.011301 .2310537
360 1 | -.1354763 .1137266 -1.19 0.234 -.3588368 .0878843
361 0 | .2387945 .0745974 3.20 0.001 .0922842 .3853047
361 1 | -.2109066 .1231149 -1.71 0.087 -.452706 .0308928
362 0 | .002561 .076416 0.03 0.973 -.1475211 .152643
362 1 | -.3449196 .1387402 -2.49 0.013 -.6174073 -.0724319
363 0 | .1106431 .0626451 1.77 0.078 -.0123926 .2336789
363 1 | -.1642916 .099942 -1.64 0.101 -.360579 .0319959
364 0 | .1120981 .0751653 1.49 0.136 -.0355276 .2597238
364 1 | -.1138568 .1051597 -1.08 0.279 -.3203919 .0926783
365 0 | .0224863 .0770132 0.29 0.770 -.1287687 .1737414
365 1 | .0063751 .1392956 0.05 0.964 -.2672033 .2799534
366 0 | .0786756 .0677887 1.16 0.246 -.0544624 .2118135
366 1 | -.1137175 .1497857 -0.76 0.448 -.4078987 .1804638
367 0 | .0472342 .0662716 0.71 0.476 -.0829241 .1773926
367 1 | -.0897144 .0820489 -1.09 0.275 -.2508596 .0714307
368 0 | 0 (omitted)
368 1 | 0 (omitted)
|
residential | -.6372152 .5779848 -1.10 0.271 -1.772385 .4979547
workplaces | -.3516455 .2013581 -1.75 0.081 -.7471155 .0438246
transit_sta~s | .0581689 .0414588 1.40 0.161 -.0232568 .1395946
parks | -.0022952 .0263351 -0.09 0.931 -.0540177 .0494272
grocery_and~y | .2286835 .1573574 1.45 0.147 -.0803686 .5377356
retail_and_~n | -.0912527 .1100078 -0.83 0.407 -.3073095 .1248042
_cons | .5892569 .7971742 0.74 0.460 -.976404 2.154918
-------------------------------------------------------------------------------

Absorbed degrees of freedom:
-----------------------------------------------------+
Absorbed FE | Categories - Redundant = Num. Coefs |
-------------+---------------------------------------|
fips | 588 588 0 *|
time | 175 0 175 |
-----------------------------------------------------+
* = FE nested within cluster; treated as redundant for DoF computation

.
end of do-file



How Can I plot the coefficient of results foe each group separately in same graph ( 0,1) groups of coefficients? in the attached results I want to plot the coefficient of regression using coeplot. you see two groups of coefficient the first one with business policy=0 and the other with business policy=1. when i plot the previous coefficient it plotted all together without any distinguish of colors between groups.

[Clyde Schechter]

I think I see the problem. The colinearity is probably arising with the absorbed effects, fips and time. As I am not entirely sure what all of these variables are, I can't be entirely sure of the details, but I suspect there may be colinearity between the variables days and time, or perhaps there is an issue of policy being constant within fips. To know in more detail, I would need to work with example data. If you want me to do that, please post back using the -dataex- command to show an example. Please be sure your example exhibits this same problem. If you are running version 17, 16 or a fully updated version 15.1 or 14.2, -dataex- is already part of your official Stata installation. If not, run -ssc install dataex- to get it. Either way, run -help dataex- to read the simple instructions for using it. -dataex- will save you time; it is easier and quicker than typing out tables. It includes complete information about aspects of the data that are often critical to answering your question but cannot be seen from tabular displays or screenshots. It also makes it possible for those who want to help you to create a faithful representation of your example to try out their code, which in turn makes it more likely that their answer will actually work in your data.

As for your other question, I do not use -coefplot- and know very little about it, so I can't help you with that. If somebody else following this thread knows the answer I hope they will chime in. If that does not happen within, say, 24 hours, I suggest starting a new thread with just that question.

[amera amery]

Thank you. As my dataset has 1 million records I just collected 1000 in the attached do file. of them and change some numbers in the code. but still have the same problem.

replace  delta_mask= 0 if mask=="2077-12-31"

summ delta_mask
g days= delta_mask - r(min)



summ days  if delta_mask == -7


* used to find the index of -7 day after shifting and 30 days after policy 
local true_neg1 = r(mean)

summ days if delta_mask== 30


g busniss_policy=business_delta_day


replace busniss_policy=1 if busniss_policy>=0

replace busniss_policy=0 if busniss_policy<0


reghdfe cases_normed i.days#busniss_policy  residential workplaces transit_stations parks grocery_and_pharmacy retail_and_recreation  if days<=36& days>=0&mask!="2077-12-31", absorb(fips time)  

margins  days#busniss_policy, post

fips: county identification number
time: time series for each county and it is the same for all counties.
I used fips and time in fixed effect regression
delta_mask: number of days from applying mask policy >0 means county applied policy <0 means did not apply the policy.
business_delta_day:number of days from applying businesspolicy >0 means county applied policy <0 means did not apply the policy.

[Clyde Schechter]

Thank you.

The data example you sent has busniss_policy = 0 in all observations, which is an additional complication that, I imagine, does not obtain in the full data set. But you should verify that. But what I think is the source of your problem is that the variable days is completely determined by the combination of time and fips. Here is the proof of the colinearity:

. regress days i.time i.fips if days<=36& days>=0&mask!="2077-12-31"

      Source |       SS           df       MS      Number of obs   =       950
-------------+----------------------------------   F(80, 869)      =         .
       Model |  107626.316        80  1345.32895   Prob > F        =         .
    Residual |           0       869           0   R-squared       =    1.0000
-------------+----------------------------------   Adj R-squared   =    1.0000
       Total |  107626.316       949  113.410238   Root MSE        =         0

------------------------------------------------------------------------------
        days | Coefficient  Std. err.      t    P>|t|     [95% conf. interval]
-------------+----------------------------------------------------------------
        time |
        172  |          1          .        .       .            .           .
        173  |          2          .        .       .            .           .
        174  |          3          .        .       .            .           .
        175  |          4          .        .       .            .           .
        176  |          5          .        .       .            .           .
        177  |          6          .        .       .            .           .
        178  |          7          .        .       .            .           .
        179  |          8          .        .       .            .           .
        180  |          9          .        .       .            .           .
        181  |         10          .        .       .            .           .
        182  |         11          .        .       .            .           .
        183  |         12          .        .       .            .           .
        184  |         13          .        .       .            .           .
        185  |         14          .        .       .            .           .
        186  |         15          .        .       .            .           .
        187  |         16          .        .       .            .           .
        188  |         17          .        .       .            .           .
        189  |         18          .        .       .            .           .
        190  |         19          .        .       .            .           .
        191  |         20          .        .       .            .           .
        192  |         21          .        .       .            .           .
        193  |         22          .        .       .            .           .
        194  |         23          .        .       .            .           .
        195  |         24          .        .       .            .           .
        196  |         25          .        .       .            .           .
        197  |         26          .        .       .            .           .
        198  |         27          .        .       .            .           .
        199  |         28          .        .       .            .           .
        200  |         29          .        .       .            .           .
        201  |         30          .        .       .            .           .
        202  |         31          .        .       .            .           .
        203  |         32          .        .       .            .           .
        204  |         33          .        .       .            .           .
        205  |         34          .        .       .            .           .
        206  |         35          .        .       .            .           .
        207  |         36          .        .       .            .           .
        208  |         37          .        .       .            .           .
        209  |         38          .        .       .            .           .
        210  |         39          .        .       .            .           .
        211  |         40          .        .       .            .           .
        212  |         41          .        .       .            .           .
        213  |         42          .        .       .            .           .
        214  |         43          .        .       .            .           .
        215  |         44          .        .       .            .           .
        216  |         45          .        .       .            .           .
        217  |         46          .        .       .            .           .
        218  |         47          .        .       .            .           .
        219  |         48          .        .       .            .           .
        220  |         49          .        .       .            .           .
        221  |         50          .        .       .            .           .
        222  |         51          .        .       .            .           .
        223  |         52          .        .       .            .           .
        224  |         53          .        .       .            .           .
        225  |         54          .        .       .            .           .
        226  |         55          .        .       .            .           .
             |
        fips |
       1003  |   8.66e-14          .        .       .            .           .
       1005  |   8.64e-14          .        .       .            .           .
       1007  |   8.64e-14          .        .       .            .           .
       1009  |   8.65e-14          .        .       .            .           .
       1011  |   8.65e-14          .        .       .            .           .
       1013  |   8.51e-14          .        .       .            .           .
       1015  |   8.64e-14          .        .       .            .           .
       1017  |   8.65e-14          .        .       .            .           .
       1019  |   8.65e-14          .        .       .            .           .
       1021  |   8.65e-14          .        .       .            .           .
       1023  |   8.65e-14          .        .       .            .           .
       1025  |   8.63e-14          .        .       .            .           .
       1027  |   8.66e-14          .        .       .            .           .
       1029  |   8.64e-14          .        .       .            .           .
       1031  |   8.64e-14          .        .       .            .           .
       1033  |   8.64e-14          .        .       .            .           .
       1035  |   8.65e-14          .        .       .            .           .
       1037  |   8.66e-14          .        .       .            .           .
       1039  |   8.43e-14          .        .       .            .           .
       1041  |   8.65e-14          .        .       .            .           .
       1043  |   8.65e-14          .        .       .            .           .
       1045  |   8.62e-14          .        .       .            .           .
       1047  |         19          .        .       .            .           .
       1049  |   8.61e-14          .        .       .            .           .
       1051  |   8.44e-14          .        .       .            .           .
             |
       _cons |        -19          .        .       .            .           .
------------------------------------------------------------------------------

Note the R² = 1. This is exact colinearity. (The 8.64e-14 coefficients are zeros with some rounding error--if you juggle these results you will see that there is an exact relationship days = time-190, except in fips 1047, where days = time - 171.) So you cannot use the days variable and also absorb both time and fips.

My first thought is to wonder why are the time and days variables so tightly related to each other (and perfectly related if we take fips into account as well). Is one of them incorrect? It is unusual to have two variables in the data set that are so strongly related to each other. Usually it's because one of them is a mistake. If they are both correct, then you are going to have to omit one of them from the model. (Or you might try removing fips instead, your data set is so large that even in the presence of such strong colinearity between days and time that you might get stable results if fips were taken out of the picture, but I doubt it since everything would then hinge on the slightly different relationship between days and time seen in fips 1047.)

[amera amery]

I am using the fixed effect regression and I have generated time (date-"2020-1-1") to be able to use it as a time series in fixed effect regression as this regression asked about time variable to be specified. my thought is that the colinearity is exist among independent variable in regression and I did not use time as independent variable in regression to get the colinearity problem. but after removing time from absorb function still get omitted for last day with buinsess_policy=1 as follows:

(dropped 92 singleton observations)
(MWFE estimator converged in 1 iterations)
note: 368.days#1.busniss_policy omitted because of collinearity

HDFE Linear regression                            Number of obs   =      3,664
Absorbing 1 HDFE group                            F(  80,   2990) =       1.48
                                                  Prob > F        =     0.0041
                                                  R-squared       =     0.5612
                                                  Adj R-squared   =     0.4624
                                                  Within R-sq.    =     0.0380
                                                  Root MSE        =     0.4730

-------------------------------------------------------------------------------
 cases_normed | Coefficient  Std. err.      t    P>|t|     [95% conf. interval]
--------------+----------------------------------------------------------------
         days#|
busniss_pol~y |
       331 1  |   -.183647    .184291    -1.00   0.319     -.544997    .1777029
       332 0  |  -.0108646   .0784753    -0.14   0.890    -.1647357    .1430065
       332 1  |  -.3288637   .1661537    -1.98   0.048    -.6546508   -.0030767
       333 0  |  -.0737479   .0779222    -0.95   0.344    -.2265345    .0790387
       333 1  |  -.3521921    .187075    -1.88   0.060    -.7190008    .0146166
       334 0  |  -.0353539   .0810662    -0.44   0.663     -.194305    .1235972
       334 1  |   -.287462   .2102734    -1.37   0.172    -.6997571    .1248331
       335 0  |  -.1327006   .0786884    -1.69   0.092    -.2869896    .0215883
       335 1  |  -.4236115   .1943978    -2.18   0.029    -.8047785   -.0424444
       336 0  |    .059865    .078678     0.76   0.447    -.0944036    .2141335
       336 1  |   -.300042   .1688129    -1.78   0.076    -.6310432    .0309592
       337 0  |  -.0118283   .0835899    -0.14   0.887    -.1757278    .1520712
       337 1  |  -.2329834   .1706959    -1.36   0.172    -.5676767    .1017098
       338 0  |   .0401325   .0794825     0.50   0.614    -.1157134    .1959783
       338 1  |   .0085674   .1757388     0.05   0.961    -.3360138    .3531486
       339 0  |   .0916985   .0810704     1.13   0.258    -.0672609     .250658
       339 1  |  -.1855247   .1708097    -1.09   0.278    -.5204412    .1493918
       340 0  |  -.0389724   .0809749    -0.48   0.630    -.1977444    .1197997
       340 1  |  -.1767039   .1684643    -1.05   0.294    -.5070217    .1536138
       341 0  |  -.0612233   .0816568    -0.75   0.453    -.2213324    .0988859
       341 1  |  -.0515542    .178749    -0.29   0.773    -.4020378    .2989293
       342 0  |   .0407878   .0803586     0.51   0.612    -.1167761    .1983516
       342 1  |  -.2141028   .1805367    -1.19   0.236    -.5680915    .1398859
       343 0  |   .0427721   .0771287     0.55   0.579    -.1084585    .1940027
       343 1  |   .0626085    .156433     0.40   0.689    -.2441187    .3693357
       344 0  |   .1319646   .0823549     1.60   0.109    -.0295134    .2934425
       344 1  |  -.1459327   .1712621    -0.85   0.394    -.4817362    .1898707
       345 0  |    .029413   .0760691     0.39   0.699    -.1197401    .1785662
       345 1  |  -.1778338   .2078138    -0.86   0.392    -.5853063    .2296387
       346 0  |   .0440757   .0784369     0.56   0.574      -.10972    .1978714
       346 1  |  -.0987653   .1760017    -0.56   0.575     -.443862    .2463314
       347 0  |    .053933   .0781458     0.69   0.490     -.099292    .2071579
       347 1  |   -.193008   .1724892    -1.12   0.263    -.5312175    .1452016
       348 0  |  -.0133792   .0798774    -0.17   0.867    -.1699995    .1432411
       348 1  |   .0940962   .1667075     0.56   0.572    -.2327768    .4209691
       349 0  |   .1277307   .0762846     1.67   0.094    -.0218449    .2773062
       349 1  |  -.2452425   .1782802    -1.38   0.169    -.5948068    .1043219
       350 0  |   .0637541    .078787     0.81   0.418    -.0907281    .2182363
       350 1  |  -.1804647   .1800949    -1.00   0.316     -.533587    .1726577
       351 0  |  -.0338998   .0792428    -0.43   0.669    -.1892758    .1214761
       351 1  |  -.4057417   .1628739    -2.49   0.013    -.7250981   -.0863854
       352 0  |   .1553104   .0798732     1.94   0.052    -.0013016    .3119225
       352 1  |  -.1738054   .1684738    -1.03   0.302    -.5041417    .1565309
       353 0  |   .0110072   .0805764     0.14   0.891    -.1469836    .1689979
       353 1  |  -.0547025   .1841086    -0.30   0.766    -.4156948    .3062898
       354 0  |   .1398289   .0792685     1.76   0.078    -.0155975    .2952553
       354 1  |  -.1390049   .1738561    -0.80   0.424    -.4798945    .2018848
       355 0  |   .0022844   .0788798     0.03   0.977    -.1523797    .1569485
       355 1  |  -.2779353   .1684323    -1.65   0.099    -.6081901    .0523195
       356 0  |   .1665103   .0794925     2.09   0.036     .0106448    .3223758
       356 1  |  -.1605142    .176079    -0.91   0.362    -.5057624     .184734
       357 0  |   .0059194   .0826253     0.07   0.943    -.1560889    .1679277
       357 1  |  -.3000363   .1795686    -1.67   0.095    -.6521269    .0520542
       358 0  |   .1036955   .0778141     1.33   0.183    -.0488791    .2562701
       358 1  |  -.1165501   .1628458    -0.72   0.474    -.4358512     .202751
       359 0  |   .1109954   .0784887     1.41   0.157    -.0429019    .2648928
       359 1  |   .0069488   .1638173     0.04   0.966    -.3142571    .3281548
       360 0  |   .0192964   .0865679     0.22   0.824    -.1504422     .189035
       360 1  |   -.078074   .1581449    -0.49   0.622    -.3881578    .2320098
       361 0  |   .1442134    .080101     1.80   0.072    -.0128453    .3012721
       361 1  |  -.1148409   .1672719    -0.69   0.492    -.4428205    .2131388
       362 0  |  -.0934121   .0840626    -1.11   0.267    -.2582385    .0714144
       362 1  |  -.1703662   .1837668    -0.93   0.354    -.5306883    .1899559
       363 0  |  -.0299809   .0828708    -0.36   0.718    -.1924704    .1325086
       363 1  |  -.1385221   .1795026    -0.77   0.440    -.4904833     .213439
       364 0  |  -.0125648   .0838218    -0.15   0.881    -.1769189    .1517894
       364 1  |   .0258663   .1679649     0.15   0.878    -.3034721    .3552047
       365 0  |  -.1051515    .079917    -1.32   0.188    -.2618494    .0515463
       365 1  |  -.0949899   .1736825    -0.55   0.584    -.4355391    .2455593
       366 0  |   .0565542    .085706     0.66   0.509    -.1114945    .2246029
       366 1  |  -.1821631   .1771512    -1.03   0.304    -.5295136    .1651875
       367 0  |  -.0518685   .0845795    -0.61   0.540    -.2177085    .1139715
       367 1  |   -.083387   .1756926    -0.47   0.635    -.4278776    .2611037
       368 0  |  -.1104242    .088316    -1.25   0.211    -.2835905     .062742
       368 1  |          0  (omitted)
              |
  residential |  -.1105294   .5132284    -0.22   0.830    -1.116846    .8957872
   workplaces |  -.2910593   .2113937    -1.38   0.169    -.7055511    .1234324
transit_sta~s |    .083278   .0464314     1.79   0.073    -.0077628    .1743188
        parks |   .0034286   .0229842     0.15   0.881    -.0416378    .0484949
grocery_and~y |    .123441   .1316366     0.94   0.348    -.1346666    .3815485
retail_and_~n |  -.0926652   .1215411    -0.76   0.446    -.3309779    .1456475
        _cons |    .099166   .7050186     0.14   0.888    -1.283205    1.481537
-------------------------------------------------------------------------------

Absorbed degrees of freedom:
-----------------------------------------------------+
 Absorbed FE | Categories  - Redundant  = Num. Coefs |
-------------+---------------------------------------|
        fips |       594           0         594     |
-----------------------------------------------------+

. 
end of do-file

do you think having . (null) values of some google mobility data may cause this problem?

Yes the data I have uploaded all Buisness_policy =0 but it is not the case in the whole data set.

[Clyde Schechter]

This is a different result and it is not a problem. Notice that this time there is only one coefficient omtited, not, as before, two. I will admit that I am puzzled why it is the last one that is omitted, since I normally expect Stata to remove the first. But that doesn't actually matter. The point is that there is no longer any problematic colinearity: what you are seeing is just the routine omission of one base category. And if you run -margins days#busniss_policy- after this, you will get estimates.

[amera amery]

The results are different because i am using the whole dataset not just the one i sent This is not the routine omission as the routine omission happened at day 331 0. so two values are omitted.

Even I tried with just keep regression with one independent variable i.days#buisness_policy but got the same problem.

[Clyde Schechter]

Well, your example data set produces very similar looking results, and -margins- does give results afterwards. Notice that while 368 1 was omitted, 368 0 was not. That's a key difference. If you still had a complete colinearity going on with the days variable, both 368 1 and 368 0 would have been omitted. Have you tried running -margins- after the regression in #9. I believe it will work.

Even I tried with just keep regression with one independent variable i.days#buisness_policy and got the same problem.

Right. This just confirms what I have asserted before: to the extent you have problems, it is with the variables days, time and fips. The other variables are not part of the problem.

[amera amery]

the regression after removing time. the days#busniss variable is an intersection variable to study days with busniss_policy.

reghdfe cases_normed i.days#busniss_policy cases_normed_delay_14days cases_normed_delay_14days_growth residential workplaces transit_stations parks grocery_and_pharmacy retail_and_recreation  if days<=368 & days>=331&mask!="2077-12-31", absorb(fips ) vce(cluster fips)

and still can not get the margin estimation as it is not estimable for all coefficient. The things I noticed that if I removed the condition mask!="2077-12-31" I will get good results without any omitted values. but I should include it in my regression as this condition mean that include the counties that applied the mask the value of mask=2077-12-31" means it did not applied the policy and it is beyond of our interest.

[amera amery]

I see this post as well on your website. They mentioned that this problem may because of lake of combination data? So when I removed the condition we will have enough data then nothing is omitted? Do you think this is the problem?!!!

In the documentation of the margins command in the Stata Base Reference Manual Release 13 you will see on page 1185 "Estimability of Margins" which discusses the problem with estimating margins when some combinations of the variables lack data.

With that said, the following will ignore the estimability issues and provide margins, leaving it to you to determine if they make sense within the limitations discussed by the cited passage.

[amera amery]

I see this answer posted on your website. They mentioned that this problem because of lake of combination data? So when I removed the condition we will have enough data then nothing is omitted? Do you think this is the problem?!!!


""In the documentation of the margins command in the Stata Base Reference Manual Release 13 you will see on page 1185 "Estimability of Margins" which discusses the problem with estimating margins when some combinations of the variables lack data.

With that said, the following will ignore the estimability issues and provide margins, leaving it to you to determine if they make sense within the limitations discussed by the cited passage.""