I am trying to test for autocorrelation in a panel data regression. ( I actually check for Newey west regression , but I could use GLS depending on the results too, to estimate my parameters) .
According with Wooldridge test for autocorrelation in panel data, I get :
but if i use actest:
So, according with Wooldridge I dont have autocorrelation but according with Cumby-Huizinga test for autocorrelation I do.
Except one test is for AR and the other for MA, why the difference in results and what should I understand out of it ?
Thank you,
Alexa
According with Wooldridge test for autocorrelation in panel data, I get :
Code:
Wooldridge test for autocorrelation in panel data
H0: no first-order autocorrelation
F( 1, 4) = 0.229
Prob > F = 0.6574
Code:
actest, lag(6) robust
Cumby-Huizinga test for autocorrelation
H0: variable is MA process up to order q
HA: serial correlation present at specified lags >q
-----------------------------------------------------------------------------
H0: q=0 (serially uncorrelated) | H0: q=specified lag-1
HA: s.c. present at range specified | HA: s.c. present at lag specified
-----------------------------------------+-----------------------------------
lags | chi2 df p-val | lag | chi2 df p-val
-----------+-----------------------------+-----+-----------------------------
1 - 1 | 49.280 1 0.0000 | 1 | 49.280 1 0.0000
1 - 2 | 70.032 2 0.0000 | 2 | 30.840 1 0.0000
1 - 3 | 74.875 3 0.0000 | 3 | 20.581 1 0.0000
1 - 4 | 76.064 4 0.0000 | 4 | 17.715 1 0.0000
1 - 5 | 77.342 5 0.0000 | 5 | 12.594 1 0.0004
1 - 6 | 78.665 6 0.0000 | 6 | 10.093 1 0.0015
-----------------------------------------------------------------------------
Test allows predetermined regressors/instruments
Test robust to heteroskedasticity
. actest , lag(12) robust
Cumby-Huizinga test for autocorrelation
H0: variable is MA process up to order q
HA: serial correlation present at specified lags >q
-----------------------------------------------------------------------------
H0: q=0 (serially uncorrelated) | H0: q=specified lag-1
HA: s.c. present at range specified | HA: s.c. present at lag specified
-----------------------------------------+-----------------------------------
lags | chi2 df p-val | lag | chi2 df p-val
-----------+-----------------------------+-----+-----------------------------
1 - 1 | 49.280 1 0.0000 | 1 | 49.280 1 0.0000
1 - 2 | 70.032 2 0.0000 | 2 | 30.840 1 0.0000
1 - 3 | 74.875 3 0.0000 | 3 | 20.581 1 0.0000
1 - 4 | 76.064 4 0.0000 | 4 | 17.715 1 0.0000
1 - 5 | 77.342 5 0.0000 | 5 | 12.594 1 0.0004
1 - 6 | 78.665 6 0.0000 | 6 | 10.093 1 0.0015
1 - 7 | 78.689 7 0.0000 | 7 | 6.005 1 0.0143
1 - 8 | 79.044 8 0.0000 | 8 | 8.566 1 0.0034
1 - 9 | 85.061 9 0.0000 | 9 | 8.026 1 0.0046
1 - 10 | 86.694 10 0.0000 | 10 | 8.178 1 0.0042
1 - 11 | 87.676 11 0.0000 | 11 | 8.010 1 0.0047
1 - 12 | 87.693 12 0.0000 | 12 | 7.864 1 0.0050
-----------------------------------------------------------------------------
Test allows predetermined regressors/instruments
Test robust to heteroskedasticity
. actest , lag(12) robust clu(_fx)
Cumby-Huizinga test for autocorrelation (Arellano-Bond)
H0: variable is MA process up to order q
HA: serial correlation present at specified lags >q
-----------------------------------------------------------------------------
H0: q=0 (serially uncorrelated) | H0: q=specified lag-1
HA: s.c. present at range specified | HA: s.c. present at lag specified
-----------------------------------------+-----------------------------------
lags | chi2 df p-val | lag | chi2 df p-val
-----------+-----------------------------+-----+-----------------------------
1 - 1 | 4.337 1 0.0373 | 1 | 4.337 1 0.0373
1 - 2 | 4.728 2 0.0941 | 2 | 3.715 1 0.0539
1 - 3 | 4.796 3 0.1874 | 3 | 3.977 1 0.0461
1 - 4 | 4.808 4* 0.3076 | 4 | 4.020 1 0.0450
1 - 5 | 5.000 5 0.4159 | 5 | 3.868 1 0.0492
1 - 6 | 5.000 6 0.5438 | 6 | 3.912 1 0.0480
1 - 7 | 5.000 7* 0.6600 | 7 | 3.585 1 0.0583
1 - 8 | 5.000 8* 0.7576 | 8 | 3.858 1 0.0495
1 - 9 | 5.000 9 0.8343 | 9 | 4.098 1 0.0429
1 - 10 | 5.000 10* 0.8912 | 10 | 4.266 1 0.0389
1 - 11 | 5.000 11 0.9312 | 11 | 4.318 1 0.0377
1 - 12 | 5.000 12* 0.9580 | 12 | 4.188 1 0.0407
-----------------------------------------------------------------------------
Test allows predetermined regressors/instruments
Test robust to heteroskedasticity and within-cluster autocorrelation
* Eigenvalues adjusted to make matrix positive semidefinite
actest, lags(6) q0 robust kernel(bartlett) bw(7)
Cumby-Huizinga test for autocorrelation
H0: variable is MA process up to order q
HA: serial correlation present at specified lags >q
-----------------------------------------------------------------------------
H0: q=0 (serially uncorrelated) | H0: q=0 (serially uncorrelated)
HA: s.c. present at range specified | HA: s.c. present at lag specified
-----------------------------------------+-----------------------------------
lags | chi2 df p-val | lag | chi2 df p-val
-----------+-----------------------------+-----+-----------------------------
1 - 1 | 4.527 1 0.0334 | 1 | 4.527 1 0.0334
1 - 2 | 9.136 2 0.0104 | 2 | 6.448 1 0.0111
1 - 3 | 11.648 3 0.0087 | 3 | 0.514 1 0.4734
1 - 4 | 11.648 4 0.0202 | 4 | 0.136 1 0.7123
1 - 5 | 12.352 5 0.0303 | 5 | 0.041 1 0.8394
1 - 6 | 12.653 6 0.0489 | 6 | 3.031 1 0.0817
-----------------------------------------------------------------------------
Test allows predetermined regressors/instruments
Test robust to heteroskedasticity
Except one test is for AR and the other for MA, why the difference in results and what should I understand out of it ?
Thank you,
Alexa
