Hello Stata community members,
I hope all of you are fine. I am working with panel dataset and I ran fixed effect by using the command xtreg and specifying the fe (options). My dataset as you can see is strongly balanced and there are no missing values (the only ones are those due to the fact that I created some lag variables. I must say that I took notes of some cross-sectional units with values far from the mean and other summary statistics. These cross-sectional units, as I am in the first stage of the analysis, are still embedded within the dataset I used to run the panel regression with fixed effects. The problem is that the coefficients are simply huge as you can see and by looking at the p-values it seems that most of the variables are not statistically significant. I still did not check for unit roots or other elements as I was shocked by the results.
Please have a look.
Coefficients are insanely huge as conf interval and std errors. I made some research and just to help you a bit I can tell you that the variables of the original dataset I have on excel are as follows:
NPL (dependent variables) in $000 which means that if it is saved on excel as 10 it equals to 10,000;
NetInterestMargin; AvgEquityAvgAssets; CosttoIncome; ROAA, LLP as %;
Assets in $000;
deltabankloans; deltaFTSEMIB; RealGDPGrowth; deltaNCLDeposits as % since they are calculated as variation in nearby years (x1-x0)/x0.
Could you please tell me why I got results as big as those that I provided you with? Is this all due to my dataset of on the other hand I did do something wrong before or during the launch of the command? Or it is due to the units of my variables?
Because I do not have any clue about it and even though I expected the results to be quite different from those I expected based on the literature, the difference is well beyond any common sense and reasonable expectation.
Thanks a lot to everybody who will try to give me some tips or support.
Maybe could sound funny to some of you, but still I will appreciate any help.
Kind regards,
Salvatore
I hope all of you are fine. I am working with panel dataset and I ran fixed effect by using the command xtreg and specifying the fe (options). My dataset as you can see is strongly balanced and there are no missing values (the only ones are those due to the fact that I created some lag variables. I must say that I took notes of some cross-sectional units with values far from the mean and other summary statistics. These cross-sectional units, as I am in the first stage of the analysis, are still embedded within the dataset I used to run the panel regression with fixed effects. The problem is that the coefficients are simply huge as you can see and by looking at the p-values it seems that most of the variables are not statistically significant. I still did not check for unit roots or other elements as I was shocked by the results.
Please have a look.
Code:
Fixed-effects (within) regression Number of obs = 1,285
Group variable: id Number of groups = 117
R-squared: Obs per group:
Within = 0.3832 min = 9
Between = 0.6398 avg = 11.0
Overall = 0.5044 max = 11
F(21,1147) = 33.93
corr(u_i, Xb) = 0.3601 Prob > F = 0.0000
-------------------------------------------------------------------------------------
NPL | Coefficient Std. err. t P>|t| [95% conf. interval]
--------------------+----------------------------------------------------------------
NetInterestMargin | 9800.016 27158.09 0.36 0.718 -43485.09 63085.13
AvgEquityAvgAssets | -4547.432 5256.832 -0.87 0.387 -14861.52 5766.653
CosttoIncome | -2946.611 812.1693 -3.63 0.000 -4540.115 -1353.107
ROAA | 358889.5 35828.86 10.02 0.000 288592.1 429187
LLP | 4.41977 .1830357 24.15 0.000 4.060647 4.778892
Assets | 1581764 1008226 1.57 0.117 -396410.9 3559938
deltabankloans | 3270.26 4493.76 0.73 0.467 -5546.651 12087.17
deltaFTSEMIB | -502.3053 675.9783 -0.74 0.458 -1828.598 823.9874
RealGDPGrowth | 8891.507 3234.141 2.75 0.006 2546.012 15237
deltaNCLDeposits | 287.1092 2509.721 0.11 0.909 -4637.049 5211.268
dummy_25 | 44582.71 76821.41 0.58 0.562 -106143.5 195309
dummy_50_75 | 64392.4 47377.17 1.36 0.174 -28563.24 157348
dummy_25_50 | 41988.99 62296.63 0.67 0.500 -80239.14 164217.1
SIZE_25_ROAA | -385542.2 43719.28 -8.82 0.000 -471320.9 -299763.5
SIZE_50_ROAA | -367153.7 60707.99 -6.05 0.000 -486264.8 -248042.5
SIZE_75_ROAA | -356867.5 48265.27 -7.39 0.000 -451565.6 -262169.4
L1_RealGDPGrowth | 11184.54 4724.096 2.37 0.018 1915.705 20453.38
L2_RealGDPGrowth | 8308.824 4608.882 1.80 0.072 -733.9618 17351.61
L1_deltaFTSEMIB | 595.1859 463.5651 1.28 0.199 -314.3447 1504.717
L1_deltabankloans | -9624.766 4660.513 -2.07 0.039 -18768.85 -480.6798
L1_deltaNCLDeposits | -7147.126 2505.654 -2.85 0.004 -12063.31 -2230.946
_cons | 237585.7 90869.38 2.61 0.009 59296.84 415874.5
--------------------+----------------------------------------------------------------
sigma_u | 302894.1
sigma_e | 261385.27
rho | .57316486 (fraction of variance due to u_i)
-------------------------------------------------------------------------------------
F test that all u_i=0: F(116, 1147) = 8.89 Prob > F = 0.0000
NPL (dependent variables) in $000 which means that if it is saved on excel as 10 it equals to 10,000;
NetInterestMargin; AvgEquityAvgAssets; CosttoIncome; ROAA, LLP as %;
Assets in $000;
deltabankloans; deltaFTSEMIB; RealGDPGrowth; deltaNCLDeposits as % since they are calculated as variation in nearby years (x1-x0)/x0.
Could you please tell me why I got results as big as those that I provided you with? Is this all due to my dataset of on the other hand I did do something wrong before or during the launch of the command? Or it is due to the units of my variables?
Because I do not have any clue about it and even though I expected the results to be quite different from those I expected based on the literature, the difference is well beyond any common sense and reasonable expectation.
Thanks a lot to everybody who will try to give me some tips or support.
Maybe could sound funny to some of you, but still I will appreciate any help.
Kind regards,
Salvatore
Comment