Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
| From | Bryon Balint <bryon.balint@belmont.edu> |
| To | "'statalist@hsphsun2.harvard.edu'" <statalist@hsphsun2.harvard.edu> |
| Subject | st: AR1 and xtgls |
| Date | Thu, 2 Jan 2014 20:05:03 +0000 |
Hello, I am using -xtgls- in Stata 13.0. My data has 2050 observations across 114 panels, which are set monthly. I also have an independent variable for a time trend, which is incremented monthly (one of the things I'm investigating is performance improvement over time). This time trend variable is also interacted with other variables. The -xtserial- test indicates that I have autocorrelation, so I reran xtgls using the c(ar1) option. My understanding of AR1 correction was that it changes the standard errors but not the coefficient estimates. The coefficients have changed, but most of them remain statistically significant. However, the coefficients on the time trend variable and all of the interactions with the time trend variable are no longer significant after using the c(ar1) option. I am trying to understand why this is the case. Is it because my time trend variable uses the same intervals (monthly) that the AR1 correction uses? Some examples below: Example 1 is without AR1 correction, Example 2 is with AR1 correction. The time trend variable is c_px_tss and its interactions are the variables following it. EXAMPLE 1 . xtgls actualval structure_yes i_structure_cost2 c_p1syn c_p1syn_cost c_variationspertask_to > tal i_c_vt_cost2 c_variationspertask_total_syn i_c_vt_syn_cost2 i_struc_c_vt i3_struc_c_vt_cos > t2 c_px_tss i_px_tss_cost2 c_px_tss_syn c_px_tss_syn_cost i_px_tss_c_vt i3_px_tss_c_vt_cost2 > i_px_tss_c_vt_syn i_px_tss_c_vt_syn_cost group2_cost newbl* monthcounter i_monthcounter_cost2, > p(h) Cross-sectional time-series FGLS regression Coefficients: generalized least squares Panels: heteroskedastic Correlation: no autocorrelation Estimated covariances = 114 Number of obs = 2050 Estimated autocorrelations = 0 Number of groups = 114 Estimated coefficients = 29 Obs per group: min = 5 avg = 17.98246 max = 25 Wald chi2(28) = 897.26 Prob > chi2 = 0.0000 ----------------------------------------------------------------------------------------------- actualval | Coef. Std. Err. z P>|z| [95% Conf. Interval] ------------------------------+---------------------------------------------------------------- structure_yes | .6757174 .2999774 2.25 0.024 .0877725 1.263662 i_structure_cost2 | .7171097 2.130786 0.34 0.736 -3.459154 4.893373 c_p1syn | -.2386655 .0818019 -2.92 0.004 -.3989943 -.0783366 c_p1syn_cost | .8814414 .5280493 1.67 0.095 -.1535163 1.916399 c_variationspertask_total | -.3886945 .1226349 -3.17 0.002 -.6290544 -.1483345 i_c_vt_cost2 | -.462433 .7868393 -0.59 0.557 -2.00461 1.079744 c_variationspertask_total_syn | -.1170541 .0399244 -2.93 0.003 -.1953044 -.0388038 i_c_vt_syn_cost2 | .6831517 .2708621 2.52 0.012 .1522717 1.214032 i_struc_c_vt | .1888914 .1344301 1.41 0.160 -.0745866 .4523695 i3_struc_c_vt_cost2 | -1.154606 .7110748 -1.62 0.104 -2.548287 .2390747 c_px_tss | .0407651 .0279762 1.46 0.145 -.0140672 .0955974 i_px_tss_cost2 | .4496724 .1982409 2.27 0.023 .0611274 .8382175 c_px_tss_syn | .02902 .0105108 2.76 0.006 .0084192 .0496208 c_px_tss_syn_cost | .0559958 .0795014 0.70 0.481 -.0998241 .2118157 i_px_tss_c_vt | .0212492 .0065342 3.25 0.001 .0084424 .0340559 i3_px_tss_c_vt_cost2 | .1660515 .0621026 2.67 0.007 .0443327 .2877703 i_px_tss_c_vt_syn | .0156071 .0049837 3.13 0.002 .0058391 .0253751 i_px_tss_c_vt_syn_cost | .0759115 .0424666 1.79 0.074 -.0073215 .1591445 _cons | 97.60483 .7827861 124.69 0.000 96.0706 99.13906 ----------------------------------------------------------------------------------------------- EXAMPLE 2 . xtgls actualval structure_yes i_structure_cost2 c_p1syn c_p1syn_cost c_variationspertask_to > tal i_c_vt_cost2 c_variationspertask_total_syn i_c_vt_syn_cost2 i_struc_c_vt i3_struc_c_vt_cos > t2 c_px_tss i_px_tss_cost2 c_px_tss_syn c_px_tss_syn_cost i_px_tss_c_vt i3_px_tss_c_vt_cost2 > i_px_tss_c_vt_syn i_px_tss_c_vt_syn_cost group2_cost newbl* monthcounter i_monthcounter_cost2, > p(h) c(ar1) force Cross-sectional time-series FGLS regression Coefficients: generalized least squares Panels: heteroskedastic Correlation: common AR(1) coefficient for all panels (0.6421) Estimated covariances = 114 Number of obs = 2050 Estimated autocorrelations = 1 Number of groups = 114 Estimated coefficients = 29 Obs per group: min = 5 avg = 17.98246 max = 25 Wald chi2(28) = 350.42 Prob > chi2 = 0.0000 ----------------------------------------------------------------------------------------------- actualval | Coef. Std. Err. z P>|z| [95% Conf. Interval] ------------------------------+---------------------------------------------------------------- structure_yes | .290638 .2755976 1.05 0.292 -.2495233 .8307993 i_structure_cost2 | -.7768766 1.751299 -0.44 0.657 -4.20936 2.655607 c_p1syn | -.0112962 .114967 -0.10 0.922 -.2366273 .214035 c_p1syn_cost | 1.510317 .7128032 2.12 0.034 .1132483 2.907386 c_variationspertask_total | -.2019513 .1342058 -1.50 0.132 -.4649898 .0610871 i_c_vt_cost2 | -1.024145 .797939 -1.28 0.199 -2.588077 .5397867 c_variationspertask_total_syn | -.0028495 .055857 -0.05 0.959 -.1123272 .1066283 i_c_vt_syn_cost2 | .3885293 .3908295 0.99 0.320 -.3774824 1.154541 i_struc_c_vt | .0365023 .1241229 0.29 0.769 -.2067742 .2797787 i3_struc_c_vt_cost2 | -.7389062 .6054729 -1.22 0.222 -1.925611 .4477989 c_px_tss | .0058479 .0399997 0.15 0.884 -.0725501 .0842458 i_px_tss_cost2 | .419579 .2525058 1.66 0.097 -.0753234 .9144813 c_px_tss_syn | .0117767 .0150597 0.78 0.434 -.0177397 .0412931 c_px_tss_syn_cost | .0067606 .1046975 0.06 0.949 -.1984427 .2119639 i_px_tss_c_vt | .0101954 .0092202 1.11 0.269 -.0078757 .0282666 i3_px_tss_c_vt_cost2 | .1195411 .0802724 1.49 0.136 -.0377898 .276872 i_px_tss_c_vt_syn | .0043021 .0071224 0.60 0.546 -.0096574 .0182617 i_px_tss_c_vt_syn_cost | .1194113 .0559825 2.13 0.033 .0096876 .229135 monthcounter | .02217 .0344892 0.64 0.520 -.0454275 .0897675 i_monthcounter_cost2 | -.0533212 .170076 -0.31 0.754 -.386664 .2800215 _cons | 98.67921 .9634511 102.42 0.000 96.79088 100.5675 ----------------------------------------------------------------------------------------------- Thanks, Dr. Bryon Balint * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/faqs/resources/statalist-faq/ * http://www.ats.ucla.edu/stat/stata/