Why is the f stat and its p-value blank when estimating regression with vce robust errors?

[Laiy Kho]

Hello,

I estimated an OLS regression with 7 independent variables and one interaction term. Among the two variables in interaction term, one is log transformed and centered and the latter is centered only. I centered the variables to reduce VIF. The model is correctly specified as per linktest results and VIF is within acceptable range as per estat vif results. I estimated the regression again with vce (robust) errors due to heteroskedasticity. However, the f stat and p value is blank. Why is that?

[Carlo Lazzaro]

Laiy:
see -help j_robustsingular-.

[Laiy Kho]

Thanks Carlo!
Actually I have noticed that this error occurs only when I add a variable that ranges from 2.3 million to 84 million. Why is it so?

Loan
Portfolio	Freq.	Percent	Cum.
2300000	370	0.04	0.04
2400000	66,820	7.01	7.05
2800000	5,350	0.56	7.61
3400000	960	0.10	7.71
4500000	20	0.00	7.71
4800000	8,860	0.93	8.64
5000000	8	0.00	8.64
7000000	4,530	0.48	9.12
7400000	29,920	3.14	12.26
7800000	12,520	1.31	13.57
7900000	930	0.10	13.67
8600000	158,192	16.59	30.26
8700000	68,300	7.16	37.43
8900000	7,480	0.78	38.21
9100000	18,440	1.93	40.15
9400000	2,164	0.23	40.37
1.09e+07	28,370	2.98	43.35
1.10e+07	450	0.05	43.40
1.21e+07	9,300	0.98	44.37
1.31e+07	2,120	0.22	44.59
1.35e+07	5,490	0.58	45.17
1.40e+07	38,080	3.99	49.17
1.47e+07	18,320	1.92	51.09
1.48e+07	1,100	0.12	51.20
1.54e+07	33,340	3.50	54.70
1.55e+07	70	0.01	54.71
1.61e+07	220	0.02	54.73
1.85e+07	157,900	16.56	71.29
2.10e+07	28,510	2.99	74.29
2.13e+07	2,470	0.26	74.54
2.41e+07	780	0.08	74.63
2.61e+07	3,220	0.34	74.96
2.77e+07	10	0.00	74.96
2.78e+07	160	0.02	74.98
3.03e+07	170	0.02	75.00
3.17e+07	106,770	11.20	86.20
3.19e+07	12,880	1.35	87.55
3.20e+07	5,610	0.59	88.14
3.29e+07	64,880	6.81	94.95
3.39e+07	480	0.05	95.00
3.43e+07	6,210	0.65	95.65
3.46e+07	7,820	0.82	96.47
4.56e+07	360	0.04	96.51
8.07e+07	20,500	2.15	98.66
8.29e+07	550	0.06	98.71
8.40e+07	12,260	1.29	100.00
Total	953,264	100.00

[Carlo Lazzaro]

Laiy:
the issue is that your model tries to estimate a number of coefficients (10) that is higher than its degrees of freedom (8).

[Laiy Kho]

Hello,

When I estimated the regression without robust error the degree of freedom reported was 10 (Image 1). Why does it change when I estimate with robust errors?

Below is the table with standard errors


std errors.PNG


Regression output with robust errors

robust.PNG

[Laiy Kho]

Hello,

When I estimated the regression without robust error the degree of freedom reported was 10 (Image 1). Why does it change when I estimate with robust errors?

Below is the table with standard errors

Regression output with robust errors

*Sorry for the repost. I had some trouble attaching the image.

[Laiy Kho]

Hello,

When I estimated the regression without robust error the degree of freedom reported was 10 (Image 1). Why does it change when I estimate with robust errors?

1. Regression output with standard errors

2. Regression output with robust errors

*Sorry for the repost. I had some trouble attaching the image.

[Carlo Lazzaro]

Laiy:
see https://blog.stata.com/2022/10/06/he...onsiderations/