I'm writing a thesis on how director reputation changes overtime. Director reputation is derived from market cap of the companies the individual director is holding seats in. I ran both an OLS and a GLS regression and get very different but significant results. I have very limited background in data science and therefore have trouble understanding which model is best suited for my research question. Any help or feedback would be greatly appreciated.
I ran a test for autocorrelation (below) and I tried running the test for heteroskedasticity but I get an error that my stata version only allows up to 800 rows. However to my understanding my data should be homoscedastic.
Code:
. xtserial DirRep YearVar
Wooldridge test for autocorrelation in panel data
H0: no first-order autocorrelation
F( 1, 6811) = 3121.669
Prob > F = 0.0000Code:
. reg DirRep YearVar
Source | SS df MS Number of obs = 50,484
-------------+---------------------------------- F(1, 50482) = 244.00
Model | 157338.357 1 157338.357 Prob > F = 0.0000
Residual | 32552654.4 50,482 644.83686 R-squared = 0.0048
-------------+---------------------------------- Adj R-squared = 0.0048
Total | 32709992.7 50,483 647.940747 Root MSE = 25.394
------------------------------------------------------------------------------
DirRep | Coefficient Std. err. t P>|t| [95% conf. interval]
-------------+----------------------------------------------------------------
YearVar | .659354 .0422111 15.62 0.000 .5766199 .7420882
_cons | 60.9483 .2171544 280.67 0.000 60.52267 61.37392
------------------------------------------------------------------------------Code:
. xtreg DirRep YearVar, re
Random-effects GLS regression Number of obs = 50,484
Group variable: DirectorID Number of groups = 8,035
R-squared: Obs per group:
Within = 0.0016 min = 1
Between = 0.0376 avg = 6.3
Overall = 0.0048 max = 10
Wald chi2(1) = 49.53
corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000
------------------------------------------------------------------------------
DirRep | Coefficient Std. err. z P>|z| [95% conf. interval]
-------------+----------------------------------------------------------------
YearVar | -.0896029 .0127312 -7.04 0.000 -.1145557 -.0646501
_cons | 61.7649 .2896927 213.21 0.000 61.19711 62.33269
-------------+----------------------------------------------------------------
sigma_u | 25.341588
sigma_e | 6.7111548
rho | .93446266 (fraction of variance due to u_i)
------------------------------------------------------------------------------Code:
* Example generated by -dataex-. For more info, type help dataex clear input double DirectorID byte YearVar int(DirYOB Network) byte(NoNed ExecVar) double DirRep 216931 1 1951 4313 1 0 61 216931 2 1951 4313 1 0 64 216931 3 1951 4313 1 0 62 216931 4 1951 4313 1 0 60 216931 5 1951 4313 1 0 65 722610 1 1939 635 1 0 96 722610 2 1939 635 1 0 97 722610 3 1939 635 1 0 97 722610 4 1939 635 1 0 97 722610 5 1939 635 1 0 97 722610 6 1939 635 1 0 96 722610 7 1939 635 1 0 97 722610 8 1939 635 1 0 96 722610 9 1939 635 1 0 96 722610 10 1939 635 1 0 96 722650 1 1945 2968 1 0 85 722650 2 1945 2968 1 0 82 722650 3 1945 2968 1 0 82 722650 4 1945 2968 1 0 82 722650 5 1945 2968 1 0 83 722650 6 1945 2968 1 0 85 722650 7 1945 2968 1 0 84 722650 8 1945 2968 1 0 85 722650 9 1945 2968 1 0 82 1050210 1 1955 1176 1 0 71 1050210 2 1955 1176 1 0 68 1050210 3 1955 1176 1 0 67 1050210 4 1955 1176 1 0 68 1050210 5 1955 1176 1 0 77 1050210 6 1955 1176 1 0 83 1050210 7 1955 1176 1 0 81 1050210 8 1955 1176 1 0 79 1050210 9 1955 1176 1 0 77 2224030 1 1950 1081 1 1 90 2224030 2 1950 1081 1 1 88 2224030 3 1950 1081 1 1 88 2224030 4 1950 1081 1 1 90 2224030 5 1950 1081 1 1 89 2224030 6 1950 1081 1 1 89 2224030 7 1950 1081 1 1 89 2224030 8 1950 1081 1 1 90 2224030 9 1950 1081 1 1 87 2224030 10 1950 1081 1 1 87 2224190 1 1954 326 1 1 52 2224190 2 1954 326 1 1 52 2224190 3 1954 326 1 1 48 2224190 4 1954 326 1 1 42 2224200 1 1947 204 1 0 52 2224200 2 1947 204 1 0 52 2224200 3 1947 204 1 0 48 2224200 4 1947 204 1 0 42 2224220 1 1951 849 2 1 84 2224220 2 1951 849 2 1 85 2224220 3 1951 849 2 1 85.5 2224220 4 1951 849 2 1 84 2224220 5 1951 849 2 1 84 2224220 6 1951 849 2 1 82 2224220 7 1951 849 2 1 81 2224220 8 1951 849 2 1 74 2224220 9 1951 849 2 1 76 2224220 10 1951 849 2 1 73 2224240 1 1938 0 1 0 70 2224240 2 1938 0 1 0 76 2224240 3 1938 0 1 0 78 2224240 4 1938 0 1 0 77 2224240 5 1938 0 1 0 79 2224240 6 1938 0 1 0 78 2224490 1 1960 1331 1 0 16 2224490 2 1960 1331 1 0 16 2224490 3 1960 1331 1 0 28.000000000000004 2224490 4 1960 1331 1 0 33 2224490 5 1960 1331 1 0 27 2224490 6 1960 1331 1 0 37 2224490 7 1960 1331 1 0 36 2224660 1 1940 152 1 0 54 2224660 2 1940 152 1 0 49 2224660 3 1940 152 1 0 36 2224660 4 1940 152 1 0 50 2224660 5 1940 152 1 0 56.00000000000001 2224660 6 1940 152 1 0 45 2224660 7 1940 152 1 0 46 2224710 1 1959 1119 1 1 90 5354000 1 1953 1405 1 0 23 5354000 2 1953 1405 1 0 15 5354000 3 1953 1405 1 0 10 5354000 4 1953 1405 1 0 5 5354570 1 1965 377 1 0 50 5354570 2 1965 377 1 0 50 5354570 3 1965 377 1 0 53 5354600 1 1959 766 1 1 44 5354600 2 1959 766 1 1 44 5354600 3 1959 766 1 1 47 5354600 4 1959 766 1 1 46 5354600 5 1959 766 1 1 48 10509380 1 1967 699 1 0 65 10509380 2 1967 699 1 0 60 11223500 1 1945 7042 1 0 92 11223500 2 1945 7042 1 0 92 11223500 3 1945 7042 1 0 92 11223500 4 1945 7042 1 0 88 end
0 Response to Different results OLS / GLS regression - how to choose the best model
Post a Comment