Sample may not include multiple panels error after trying to perform Breusch Godfrey test for

[Manuel Ekman]

Dear Everyone,

I am using panel data and I want to test for autocorrelation in my dataset. The data covers 1059 companies, with for each company 2 (DV) Dependent Variables (ROA and Tobin's Q) and 5 (IV) Independent Variables. The data covers a maximum of 81 months (Jan 2014 - Oct 2020), where most of the companies only cover 24-36 months. Furthermore, in some cases there is data missing in the DV or IV in the relevant month (see the #N/A in the overview). Overview:

ExtractionDate	EntityName	DV1	DV2	IV1	IV2	IV3	IV4	IV5	CompanyID
1-9-2014	Company A	#N/A	0,128	6,296	0,9	2536,192	30,308	0,004	1
1-10-2014	Company A	#N/A	0,128	6,296	0,9	2536,192	30,308	0,004	1
1-11-2014	Company A	#N/A	0,128	6,296	0,9	2536,192	30,308	0,004	1
1-12-2014	Company A	#N/A	0,128	6,296	0,9	2536,192	30,308	0,004	1
1-1-2015	Company A	5,158	0,079	6,927	0,9	892,200	0,000	0,195	1
1-2-2015	Company A	5,158	0,079	6,927	0,9	892,200	0,000	0,195	1
1-3-2015	Company A	5,158	0,079	6,927	0,9	892,200	0,000	0,195	1
1-4-2015	Company A	5,158	0,079	6,927	0,9	892,200	0,000	0,195	1
1-5-2015	Company A	5,158	0,079	6,927	0,9	892,200	0,000	0,195	1
1-11-2015	Company B	5,491	0,135	8,580	0,9	1107,4	0	0,170	2
1-12-2015	Company B	5,491	0,135	8,580	0,9	1107,4	0	0,170	2
1-1-2016	Company B	4,229	0,106	8,616	0,9	1399	0	0,181	2
1-2-2016	Company B	4,229	0,106	8,616	0,9	1399	0	0,181	2
1-3-2016	Company B	4,229	0,106	8,616	0,9	1399	0	0,181	2
1-4-2016	Company B	4,229	0,106	8,616	0,9	1399	0	0,181	2
1-5-2016	Company B	4,229	0,106	8,616	0,9	1399	0	0,181	2
1-6-2016	Company B	4,229	0,106	8,616	0,9	1399	0	0,181	2
1-1-2014	Company C	0,946	0,00713573	14,755008	1,15	870128	352	0,0539174	3
1-2-2014	Company C	0,946	0,00713573	14,755008	1,15	870128	352	0,0539174	3
1-3-2014	Company C	0,946	0,00713573	14,755008	1,15	870128	352	0,0539174	3
1-4-2014	Company C	0,946	0,00713573	14,755008	1,15	870128	352	0,0539174	3
1-5-2014	Company C	0,946	0,00713573	14,755008	1,15	870128	352	0,0539174	3
1-6-2014	Company C	0,946	0,00713573	14,755008	1,15	870128	352	0,0539174	3
1-7-2014	Company C	0,946	0,00713573	14,755008	1,15	870128	352	0,0539174	3
1-8-2014	Company C	0,809	0,00229936	14,755008	1,15	930695	352	0,0635052	3
1-9-2014	Company C	0,809	0,00229936	14,755008	1,15	930695	352	0,0635052	3
1-10-2014	Company C	0,809	0,00229936	14,755008	1,15	930695	352	0,0635052	3
1-11-2014	Company C	0,809	0,00229936	14,755008	1,15	930695	352	0,0635052	3

As you can see in the overview above I have generated a CompanyID to sort and group the panel data per EntityName and ExtractionDate, using the code:
sort EntityName ExtractionDate
egen CompanyID = group(EntityName)

To follow up on the latter, I have stated to STATA that I use panel data using the commands:
xtset CompanyID ExtractionDate
tsset CompanyID ExtractionDate
(resulting in):

. xtset CompanyID ExtractionDate
panel variable: CompanyID (unbalanced)
time variable: ExtractionD~e, 1/1/2014 to 10/1/2020, but with gaps
delta: 1 day

. tsset CompanyID ExtractionDate
panel variable: CompanyID (unbalanced)
time variable: ExtractionD~e, 1/1/2014 to 10/1/2020, but with gaps
delta: 1 day

I want to test for autocorrelation by using the Breusch-Godfrey test (command: estat bgodfrey), but whatever I try I keep getting the error 'sample may not include multiple panels'. The commands I use are:
reg DV1 IV1 IV2 IV3 IV4 IV5
estat bgodfrey
(and then I receive the error 'sample may not include multiple panels')

Does anyone know how I can avoid this error and perform a Breusch-Godfrey test?

[Carlo Lazzaro]

Manuel:
welcome to this forum.
The community-contributed command to test for autocorrelation after -xtreg- is named -xtserial- (just type -search xtserial- to spot and install it);

[Manuel Ekman]

Hi Carlo,

Thank you for your reply.

I have tried the xtserial command but receive the 'no observation r(2000)' error

If I remember correctly I need to search for duplicates and tried the 'duplicates report' command:
duplicates report DV1 IV1 IV2 IV3 IV4 IV5
(received the following result):

copies | observations surplus
----------+---------------------------
1 | 3765 0
2 | 2852 1426
3 | 2820 1880
4 | 3252 2439
5 | 2320 1856
6 | 2292 1910
7 | 2044 1752
8 | 1688 1477
9 | 1503 1336
10 | 1540 1386
11 | 8767 7970
12 | 25296 23188
15 | 15 14
24 | 24 23
287 | 287 286
--------------------------------------

Do you know how I can fix this?

Thank you in advance

Kind regards,
Manuel

[Carlo Lazzaro]

Manuel:
try -duplicates drop- if yiou want to keep the first observation only for each group of duplicated observations.

[Manuel Ekman]

I have tried the command -duplicates drop- and get:

Duplicates in terms of all variables

(0 observations are duplicates)

I still cannot perform a xtserial test as it gives me the following error:

no observations
r(2000);

What am I doing wrong?

[Carlo Lazzaro]

Manuel:
could you please provide an example/excerpt of your data via -dataex-? Thanks.

[Manuel Ekman]

Carlo:

See -dataex- output below:

* Example generated by -dataex-. To install: ssc install dataex
clear
input double ROAWinsor1 float(ESGP lnSize Beta) double(Debt RandD) float CompanyID int ExtractionDate
.07738895750423128 28.34461 8.565343 .9 .000050889712272997575 0 1 21185
.07738895750423128 28.34461 8.565343 .9 .000050889712272997575 0 1 21216
.07738895750423128 28.34461 8.565343 .9 .000050889712272997575 0 1 21244
.07738895750423128 28.34461 8.565343 .9 .000050889712272997575 0 1 21275
.07738895750423128 28.34461 8.565343 .9 .000050889712272997575 0 1 21305
.07738895750423128 28.34461 8.565343 .9 .000050889712272997575 0 1 21336
.07738895750423128 28.34461 8.565343 .9 .000050889712272997575 0 1 21366
.07738895750423128 28.34461 8.565343 .9 .000050889712272997575 0 1 21397
.07738895750423128 28.34461 8.565343 .9 .000050889712272997575 0 1 21428
.07738895750423128 28.34461 8.565343 .9 .000050889712272997575 0 1 21458
.07738895750423128 28.34461 8.565343 .9 .000050889712272997575 0 1 21489
.07738895750423128 . 8.565343 . .000050889712272997575 0 1 21519
.05781851405707013 . 8.773685 . .007395297761788094 0 1 21550
.05781851405707013 . 8.773685 . .007395297761788094 0 1 21581
.05781851405707013 27.22727 8.773685 .9337838 .007395297761788094 0 1 21609
.05781851405707013 27.22727 8.773685 .9337838 .007395297761788094 0 1 21640
.05781851405707013 27.22727 8.773685 .9337838 .007395297761788094 0 1 21670
.05781851405707013 27.22727 8.773685 .9337838 .007395297761788094 0 1 21701
.05781851405707013 27.257267 8.773685 .9337838 .007395297761788094 0 1 21731
.05781851405707013 27.257267 8.773685 .9337838 .007395297761788094 0 1 21762
.05781851405707013 27.257267 8.773685 .9337838 .007395297761788094 0 1 21793
.05781851405707013 27.257267 8.773685 .9337838 .007395297761788094 0 1 21823
.05781851405707013 27.257267 8.773685 .9337838 .007395297761788094 0 1 21854
.05781851405707013 25.625067 8.773685 .8702703 .007395297761788094 0 1 21884
. 25.70614 . .8702703 . . 1 21915
. 25.37494 . .8702703 . . 1 21946
. 25.37494 . .8702703 . . 1 21975
. 25.37494 . .8702703 . . 1 22006
. 25.22494 . .8702703 . . 1 22036

[Carlo Lazzaro]

Manuel:
type:

format %td ExtractionDate

and then re-run the -xtset- and -xtserial- commands.

[Manuel Ekman]

Carlo:

I have used the 'format%td ExtractionDate' command, but unfortunately after re-running the -xtset- and -xtserial- commands it still does not work. Again STATA gives the output:

. format %td ExtractionDate

. xtset CompanyID ExtractionDate
panel variable: CompanyID (unbalanced)
time variable: ExtractionD~e, 01jan2014 to 01oct2020, but with gaps
delta: 1 day

. xtserial TobinsQWinsor1 ESGP Size Beta Debt RandD
no observations
r(2000);

[Carlo Lazzaro]

Manuel:
have you double-checked that you have more than one -panelid-?

[Manuel Ekman]

Carlo:

I believe I have multiple -panel_id-.

I generated a -CompanyID- as seen in my first post to group the data per company. This is because the analysis I want to conduct is on how -ESGP- (IV) affects -TobinsQ- (DV). I created the -panel_id- the following way:
sort EntityName ExtractionDate
egen CompanyID = group(EntityName)

I go more in depth in my first post to how I constructed the data.

Do I need to construct a -panel_id- another way?

Thank you in advance again for your help Carlo!

[Carlo Lazzaro]

Manuel:
-xtserial- works by firts-differencing variables. Double-check that your dataset supports this approach.
As an example, there's no gain in differencing a constant.