Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Re: Fixed-Effect Model versus Random-Effect Model

    I examine the deterrent effect of executions on homicides between 1979 and 1998 in the United States, by using the panel data. Because the state-day is my unit of analysis, my data set is simply large with 352,555 cases. And, the total number of variables is around 195. For your further information, I use the negative binomial regression model (xtnbreg) because the dependent variable is a count data and the conditional mean is greater than the conditional variance.

    What makes me confused about the results of the Stata analysis is this: While my fixed-effect model can converge, my random-effect model cannot converge. I did not change the equation model between the fixed-effect model and the random-effect model. So, I used the maxiter command in order to estimate the random-effect model. And, the Hausman test shows that the random-effect model is more appropriate than the fixed-effect model.

    Anyone should be appreciated if you can let me know why this happens and whether both fixed-effect and random-effect models can converge at the same time.
    Last edited by Moonki Hong; 27 Oct 2015, 13:46.

  • #2
    I can't give you a definitive answer to your questions based on the information given, but here is some general advice that may be helpful.

    1. The fixed and random effects estimators are quite different, and it is not unusual to have difficulty getting random effects estimators to converge.

    2. A common (but definitely only one of many) cause of non-convergence of random effects estimators is that one or more of the variance components is close to zero. If you look at the output you got with the -maxiter- option, you will probably be able to see if that is the problem in your case (assuming you set -maxiter- large enough that you were clearly out of the transient and into the "hung up" phase of estimation). If you were to show us the exact command and Stata's output (by copying from your log file or the Results window into a code block--see FAQ), someone might be able to give you better advice on why your analysis is not converging and what to do about it.

    3. The results you get from terminating a non-converging estimation with -maxiter- do not actually represent the correct estimates of any model. They are not usable, except possibly to diagnose the cause of non-convergence. In particular, you cannot rely on a Hausman test that uses results that were prematurely terminated with -maxiter-. Those results can be quite far from the correct estimates for the random effects model. You are, in effect, using Hausman to test whether your fixed-effects model is better than some non-model!

    Comment


    • #3
      I really appreciate you, Clyde, for your advice.
      I attach the command and Stata's output as follows. My FE model is exactly same as my RE model, except for the "fe" option at the end of the command.

      . set maxiter 12

      . xtnbreg HOMICIDE_COUNTS E_B_14-E_B_1 EXECUTE_DATE E_A_1-E_A_14 NEWS_B_14-NEWS_B_1 NEWS_ON_THE_GIVEN_EXECUTION NEWS_A_1-NEWS_A_14 TV_B_14-TV_B_1 TV_ON_THE_GIVEN_EXECUTION TV_A_1-TV_A_14 ACTIVE_DEATH_PENALTY DUMMY80-DUMMY98 feb-dec SUN_DUMMY TUE_DUMMY-SAT_DUMMY count NEW_YEAR NEW_YEAR_LAG_1 GOOD_FRIDAY-CHRISTMAS_LAG_1 STATE_RESIDENTS STATE_PRISONERS STATE_PRISONER_RATE alabama alaska arkansas colorado connecticut delaware washington_dc florida georgia hawaii idaho illinois indiana iowa kansas kentucky louisiana maine maryland massachusetts michigan minnesota mississippi missouri montana nebraska nevada new_hampshire new_jersey new_mexico new_york north_dakota ohio oklahoma oregon pennsylvania rhode_island south_carolina south_dakota tennessee texas utah vermont virginia washington west_virginia wisconsin wyoming, re
      note: TV_B_14 omitted because of collinearity
      note: TV_B_13 omitted because of collinearity
      note: TV_B_12 omitted because of collinearity
      note: TV_B_11 omitted because of collinearity
      note: TV_B_10 omitted because of collinearity
      note: TV_B_9 omitted because of collinearity
      note: TV_B_8 omitted because of collinearity
      note: TV_B_7 omitted because of collinearity
      note: TV_A_4 omitted because of collinearity
      note: TV_A_5 omitted because of collinearity
      note: TV_A_6 omitted because of collinearity
      note: TV_A_7 omitted because of collinearity
      note: TV_A_8 omitted because of collinearity
      note: TV_A_9 omitted because of collinearity
      note: TV_A_10 omitted because of collinearity
      note: TV_A_11 omitted because of collinearity
      note: TV_A_12 omitted because of collinearity
      note: TV_A_13 omitted because of collinearity
      note: TV_A_14 omitted because of collinearity

      Fitting negative binomial (constant dispersion) model:

      Iteration 0: log likelihood = -1047554.8
      Iteration 1: log likelihood = -601510.82
      Iteration 2: log likelihood = -423630.05
      Iteration 3: log likelihood = -405776.02
      Iteration 4: log likelihood = -405619.18
      Iteration 5: log likelihood = -405619.05
      Iteration 6: log likelihood = -405619.05

      Iteration 0: log likelihood = -570002.9
      Iteration 1: log likelihood = -556642.09
      Iteration 2: log likelihood = -551817.09
      Iteration 3: log likelihood = -551813.02
      Iteration 4: log likelihood = -551813.02

      Iteration 0: log likelihood = -551813.02 (not concave)
      Iteration 1: log likelihood = -541885.58 (not concave)
      Iteration 2: log likelihood = -521847.62 (not concave)
      Iteration 3: log likelihood = -472295.36
      Iteration 4: log likelihood = -453343.91
      Iteration 5: log likelihood = -407705.99 (not concave)
      Iteration 6: log likelihood = -406579.11
      Iteration 7: log likelihood = -403443.42
      Iteration 8: log likelihood = -403305.47
      Iteration 9: log likelihood = -403304.58
      Iteration 10: log likelihood = -403304.58

      Fitting full model:

      Iteration 0: log likelihood = -417754.68
      Iteration 1: log likelihood = -405308.66 (not concave)
      Iteration 2: log likelihood = -405228.39 (not concave)
      Iteration 3: log likelihood = -405227.12 (not concave)
      Iteration 4: log likelihood = -405226.82 (not concave)
      Iteration 5: log likelihood = -405226.78 (not concave)
      Iteration 6: log likelihood = -405226.75 (not concave)
      Iteration 7: log likelihood = -405226.72 (not concave)
      Iteration 8: log likelihood = -405226.69 (not concave)
      Iteration 9: log likelihood = -405226.65 (not concave)
      Iteration 10: log likelihood = -405226.62 (not concave)
      Iteration 11: log likelihood = -405226.59 (not concave)
      Iteration 12: log likelihood = -405226.55 (not concave)
      convergence not achieved

      Random-effects negative binomial regression Number of obs = 372,527
      Group variable: REF_STATE Number of groups = 51

      Random effects u_i ~ Beta Obs per group:
      min = 7,291
      avg = 7,304.5
      max = 7,305

      Wald chi2(173) = 52270.08
      Log likelihood = -405226.55 Prob > chi2 = 0.0000

      ---------------------------------------------------------------------------------------------
      HOMICIDE_COUNTS | Coef. Std. Err. z P>|z| [95% Conf. Interval]
      ----------------------------+----------------------------------------------------------------
      E_B_14 | -.028741 .0331953 -0.87 0.387 -.0938026 .0363207
      E_B_13 | -.0589608 .0327311 -1.80 0.072 -.1231126 .005191
      E_B_12 | -.0492133 .0315537 -1.56 0.119 -.1110573 .0126308
      E_B_11 | .0053078 .0303148 0.18 0.861 -.0541082 .0647238
      E_B_10 | -.0168103 .0305119 -0.55 0.582 -.0766125 .0429919
      E_B_9 | .0068587 .0313018 0.22 0.827 -.0544918 .0682092
      E_B_8 | .0100451 .0324991 0.31 0.757 -.053652 .0737422
      E_B_7 | -.0013992 .0328156 -0.04 0.966 -.0657167 .0629182
      E_B_6 | -.0012705 .0319074 -0.04 0.968 -.0638079 .0612669
      E_B_5 | .0022181 .0309684 0.07 0.943 -.058479 .0629151
      E_B_4 | -.0543667 .0311822 -1.74 0.081 -.1154827 .0067493
      E_B_3 | -.019812 .0307057 -0.65 0.519 -.0799941 .0403702
      E_B_2 | -.0325084 .0321712 -1.01 0.312 -.0955629 .030546
      E_B_1 | -.0449667 .0337889 -1.33 0.183 -.1111918 .0212584
      EXECUTE_DATE | -.0578468 .0344912 -1.68 0.094 -.1254483 .0097548
      E_A_1 | -.0264862 .0331603 -0.80 0.424 -.0914792 .0385069
      E_A_2 | -.0105916 .0312482 -0.34 0.735 -.071837 .0506539
      E_A_3 | -.028007 .0309268 -0.91 0.365 -.0886224 .0326084
      E_A_4 | -.0378161 .030883 -1.22 0.221 -.0983458 .0227136
      E_A_5 | -.1076913 .0329235 -3.27 0.001 -.1722202 -.0431625
      E_A_6 | -.0684121 .0335844 -2.04 0.042 -.1342363 -.0025879
      E_A_7 | -.0464334 .0333214 -1.39 0.163 -.1117421 .0188754
      E_A_8 | .015795 .0315156 0.50 0.616 -.0459743 .0775644
      E_A_9 | -.0238829 .0313661 -0.76 0.446 -.0853593 .0375935
      E_A_10 | -.0080665 .0304418 -0.26 0.791 -.0677314 .0515984
      E_A_11 | -.0197469 .0304831 -0.65 0.517 -.0794926 .0399988
      E_A_12 | .0342324 .0308773 1.11 0.268 -.026286 .0947508
      E_A_13 | -.0495114 .0332508 -1.49 0.136 -.1146817 .015659
      E_A_14 | -.0525618 .0333839 -1.57 0.115 -.1179931 .0128694
      NEWS_B_14 | -.0680554 .0846898 -0.80 0.422 -.2340444 .0979335
      NEWS_B_13 | .0712707 .0521215 1.37 0.172 -.0308856 .173427
      NEWS_B_12 | .0077817 .0456306 0.17 0.865 -.0816526 .0972161
      NEWS_B_11 | -.0228186 .0345004 -0.66 0.508 -.0904381 .0448009
      NEWS_B_10 | -.0036898 .0517029 -0.07 0.943 -.1050257 .0976461
      NEWS_B_9 | .0208582 .030438 0.69 0.493 -.0387992 .0805155
      NEWS_B_8 | .0206577 .0281825 0.73 0.464 -.0345789 .0758943
      NEWS_B_7 | .0060454 .0305971 0.20 0.843 -.0539238 .0660147
      NEWS_B_6 | .0115917 .0224771 0.52 0.606 -.0324626 .0556459
      NEWS_B_5 | .0024457 .0229452 0.11 0.915 -.0425261 .0474175
      NEWS_B_4 | .002871 .0177713 0.16 0.872 -.03196 .0377021
      NEWS_B_3 | .0370173 .0246765 1.50 0.134 -.0113478 .0853823
      NEWS_B_2 | -.0206223 .0167585 -1.23 0.218 -.0534684 .0122237
      NEWS_B_1 | -.0003801 .0107215 -0.04 0.972 -.0213939 .0206337
      NEWS_ON_THE_GIVEN_EXECUTION | .0063272 .0066685 0.95 0.343 -.0067429 .0193973
      NEWS_A_1 | -.0012792 .0040039 -0.32 0.749 -.0091267 .0065682
      NEWS_A_2 | .0016245 .0103563 0.16 0.875 -.0186736 .0219225
      NEWS_A_3 | -.0327944 .0197997 -1.66 0.098 -.0716011 .0060123
      NEWS_A_4 | -.0196253 .0245694 -0.80 0.424 -.0677803 .0285298
      NEWS_A_5 | -.0026384 .0234808 -0.11 0.911 -.0486599 .0433832
      NEWS_A_6 | .0340585 .0491907 0.69 0.489 -.0623534 .1304705
      NEWS_A_7 | .0422498 .0622162 0.68 0.497 -.0796916 .1641912
      NEWS_A_8 | .0586691 .0816751 0.72 0.473 -.1014112 .2187493
      NEWS_A_9 | -.0605562 .0655971 -0.92 0.356 -.1891242 .0680117
      NEWS_A_10 | .1971775 .0678718 2.91 0.004 .0641513 .3302037
      NEWS_A_11 | .0777717 .0766122 1.02 0.310 -.0723854 .2279288
      NEWS_A_12 | -.0974913 .077445 -1.26 0.208 -.2492806 .054298
      NEWS_A_13 | .1597103 .2829284 0.56 0.572 -.3948193 .7142399
      NEWS_A_14 | .1675256 .1201547 1.39 0.163 -.0679731 .4030244
      TV_B_14 | 0 (omitted)
      TV_B_13 | 0 (omitted)
      TV_B_12 | 0 (omitted)
      TV_B_11 | 0 (omitted)
      TV_B_10 | 0 (omitted)
      TV_B_9 | 0 (omitted)
      TV_B_8 | 0 (omitted)
      TV_B_7 | 0 (omitted)
      TV_B_6 | -.0440536 .0921273 -0.48 0.633 -.2246197 .1365126
      TV_B_5 | .0410851 .073504 0.56 0.576 -.1029802 .1851503
      TV_B_4 | -.1964013 .1563277 -1.26 0.209 -.5027979 .1099953
      TV_B_3 | -.0579082 .0530088 -1.09 0.275 -.1618036 .0459872
      TV_B_2 | -.0116014 .0265489 -0.44 0.662 -.0636363 .0404335
      TV_B_1 | .0208736 .0144261 1.45 0.148 -.007401 .0491482
      TV_ON_THE_GIVEN_EXECUTION | -.0040905 .0116666 -0.35 0.726 -.0269566 .0187755
      TV_A_1 | -.0115414 .0355595 -0.32 0.746 -.0812366 .0581539
      TV_A_2 | .0689328 .1060879 0.65 0.516 -.1389957 .2768612
      TV_A_3 | .0330499 .051299 0.64 0.519 -.0674943 .1335941
      TV_A_4 | 0 (omitted)
      TV_A_5 | 0 (omitted)
      TV_A_6 | 0 (omitted)
      TV_A_7 | 0 (omitted)
      TV_A_8 | 0 (omitted)
      TV_A_9 | 0 (omitted)
      TV_A_10 | 0 (omitted)
      TV_A_11 | 0 (omitted)
      TV_A_12 | 0 (omitted)
      TV_A_13 | 0 (omitted)
      TV_A_14 | 0 (omitted)
      ACTIVE_DEATH_PENALTY | -.2313651 .0136983 -16.89 0.000 -.2582133 -.2045169
      DUMMY80 | .065192 .0106629 6.11 0.000 .044293 .086091
      DUMMY81 | .0338013 .0107439 3.15 0.002 .0127437 .054859
      DUMMY82 | -.0328875 .0109424 -3.01 0.003 -.0543342 -.0114407
      DUMMY83 | -.1347029 .0112454 -11.98 0.000 -.1567435 -.1126623
      DUMMY84 | -.164686 .0113594 -14.50 0.000 -.1869499 -.142422
      DUMMY85 | -.1639111 .011396 -14.38 0.000 -.1862469 -.1415754
      DUMMY86 | -.0775256 .0112198 -6.91 0.000 -.099516 -.0555351
      DUMMY87 | -.1115763 .0114032 -9.78 0.000 -.1339261 -.0892264
      DUMMY88 | -.0781959 .0114183 -6.85 0.000 -.1005754 -.0558164
      DUMMY89 | -.0432522 .0115521 -3.74 0.000 -.0658938 -.0206106
      DUMMY90 | .0409036 .0116186 3.52 0.000 .0181317 .0636756
      DUMMY91 | .1115431 .0116453 9.58 0.000 .0887187 .1343675
      DUMMY92 | .0606206 .0120624 5.03 0.000 .0369788 .0842623
      DUMMY93 | .08626 .0122284 7.05 0.000 .0622928 .1102273
      DUMMY94 | .0555243 .012689 4.38 0.000 .0306542 .0803943
      DUMMY95 | -.0092212 .0133853 -0.69 0.491 -.0354558 .0170135
      DUMMY96 | -.0726742 .0139263 -5.22 0.000 -.0999692 -.0453792
      DUMMY97 | -.1193252 .0141596 -8.43 0.000 -.1470775 -.091573
      DUMMY98 | -.2036439 .0145729 -13.97 0.000 -.2322062 -.1750815
      feb | .0090779 .0088427 1.03 0.305 -.0082534 .0264092
      mar | -.0082663 .0086972 -0.95 0.342 -.0253126 .0087799
      apr | -.0078622 .0089843 -0.88 0.382 -.0254711 .0097466
      may | .0033951 .0087825 0.39 0.699 -.0138182 .0206084
      jun | .0458825 .0086417 5.31 0.000 .028945 .0628199
      jul | .0963466 .0086064 11.19 0.000 .0794784 .1132148
      aug | .1188549 .0084347 14.09 0.000 .1023232 .1353867
      sep | .0698021 .0087393 7.99 0.000 .0526734 .0869308
      oct | .0454384 .0085822 5.29 0.000 .0286177 .0622591
      nov | .0195918 .0088408 2.22 0.027 .0022642 .0369194
      dec | .0354955 .0087478 4.06 0.000 .01835 .0526409
      SUN_DUMMY | .2295993 .0065071 35.28 0.000 .2168457 .2423529
      TUE_DUMMY | -.0293927 .0069308 -4.24 0.000 -.0429769 -.0158086
      WED_DUMMY | -.0436006 .0069218 -6.30 0.000 -.057167 -.0300341
      THUR_DUMMY | -.0285432 .0069354 -4.12 0.000 -.0421363 -.0149501
      FRI_DUMMY | .0818981 .0067793 12.08 0.000 .068611 .0951852
      SAT_DUMMY | .3204084 .0064049 50.03 0.000 .307855 .3329618
      count | .0086504 .0126985 0.68 0.496 -.0162381 .0335389
      NEW_YEAR | .6340795 .0254486 24.92 0.000 .5842011 .6839579
      NEW_YEAR_LAG_1 | .016122 .0338098 0.48 0.633 -.050144 .0823881
      GOOD_FRIDAY | .077672 .034183 2.27 0.023 .0106745 .1446695
      GOODFRIDAY_LAG_1 | -.0085628 .0311174 -0.28 0.783 -.0695517 .0524262
      easter | -.0173417 .0328495 -0.53 0.598 -.0817256 .0470421
      EASTER_LAG_1 | -.0423568 .0367039 -1.15 0.248 -.1142951 .0295816
      MEMORIAL_DAY | .0415041 .035019 1.19 0.236 -.0271319 .1101401
      MEMORIAL_LAG_1 | .063259 .0350291 1.81 0.071 -.0053968 .1319149
      INDEPENDENCE_DAY | .2126124 .0293059 7.25 0.000 .1551739 .270051
      INDEPENDENCE_LAG_1 | .0668095 .031228 2.14 0.032 .0056038 .1280152
      LABOR_DAY | .1665686 .0319985 5.21 0.000 .1038527 .2292846
      LABOR_LAG_1 | .0847011 .0337805 2.51 0.012 .0184926 .1509096
      thanksgiving | .1771112 .0329725 5.37 0.000 .1124863 .2417361
      THANKSGIVING_LAG_1 | .063462 .0329515 1.93 0.054 -.0011217 .1280457
      christmas | .1589992 .0311661 5.10 0.000 .0979148 .2200836
      CHRISTMAS_LAG_1 | -.0801671 .0344045 -2.33 0.020 -.1475987 -.0127355
      STATE_RESIDENTS | 9.99e-08 1.95e-09 51.26 0.000 9.61e-08 1.04e-07
      STATE_PRISONERS | -8.08e-06 1.91e-07 -42.30 0.000 -8.46e-06 -7.71e-06
      STATE_PRISONER_RATE | .0004478 .0000322 13.91 0.000 .0003847 .0005109
      alabama | .0307548 .0441154 0.70 0.486 -.0557098 .1172195
      alaska | -2.228478 .0601329 -37.06 0.000 -2.346336 -2.110619
      arkansas | -.5240389 .0469557 -11.16 0.000 -.6160704 -.4320074
      colorado | -.7125648 .0478744 -14.88 0.000 -.8063969 -.6187327
      connecticut | -.9129528 .0486342 -18.77 0.000 -1.008274 -.8176315
      delaware | -2.12435 .0602404 -35.26 0.000 -2.242419 -2.006281
      washington_dc | -.7795323 .0545861 -14.28 0.000 -.8865191 -.6725456
      florida | .4011227 .0407327 9.85 0.000 .3212881 .4809574
      georgia | .2938617 .0424901 6.92 0.000 .2105827 .3771408
      hawaii | -2.066226 .0603295 -34.25 0.000 -2.18447 -1.947982
      idaho | -2.270276 .0611803 -37.11 0.000 -2.390188 -2.150365
      illinois | .3145361 .0410994 7.65 0.000 .2339827 .3950895
      indiana | -.386753 .0441879 -8.75 0.000 -.4733596 -.3001463
      iowa | -1.998452 .0570984 -35.00 0.000 -2.110363 -1.886541
      kansas | -1.245789 .0512404 -24.31 0.000 -1.346219 -1.14536
      kentucky | -.5548029 .0464612 -11.94 0.000 -.6458651 -.4637407
      louisiana | .3145962 .0433701 7.25 0.000 .2295923 .3996
      maine | -2.552571 .0655072 -38.97 0.000 -2.680963 -2.42418
      maryland | -.0746866 .0437723 -1.71 0.088 -.1604788 .0111056
      massachusetts | -1.045711 .0481667 -21.71 0.000 -1.140116 -.9513059
      michigan | .0601897 .043879 1.37 0.170 -.0258116 .146191
      minnesota | -1.444558 .0525367 -27.50 0.000 -1.547528 -1.341588
      mississippi | -.1555252 .0456611 -3.41 0.001 -.2450192 -.0660311
      missouri | -.0612597 .0435626 -1.41 0.160 -.1466408 .0241215
      montana | -1.985549 .0601946 -32.99 0.000 -2.103528 -1.867569
      nebraska | -1.719687 .0566902 -30.33 0.000 -1.830798 -1.608576
      nevada | -.8800762 .0502699 -17.51 0.000 -.9786034 -.7815491
      new_hampshire | -2.756049 .0677355 -40.69 0.000 -2.888808 -2.623289
      new_jersey | -.4909603 .0436043 -11.26 0.000 -.5764232 -.4054974
      new_mexico | -.7837454 .0493282 -15.89 0.000 -.8804269 -.6870638
      new_york | .0682637 .0425936 1.60 0.109 -.0152183 .1517456
      north_dakota | -3.272866 .077212 -42.39 0.000 -3.424199 -3.121534
      ohio | -.2856764 .0420944 -6.79 0.000 -.3681799 -.203173
      oklahoma | -.6734823 .0438681 -15.35 0.000 -.7594622 -.5875024
      oregon | -1.029692 .0496269 -20.75 0.000 -1.126959 -.9324247
      pennsylvania | -.2361486 .0417255 -5.66 0.000 -.3179291 -.1543681
      rhode_island | -2.068152 .0606995 -34.07 0.000 -2.187121 -1.949184
      south_carolina | -.2485304 .0453299 -5.48 0.000 -.3373755 -.1596854
      south_dakota | -2.436258 .0658662 -36.99 0.000 -2.565353 -2.307162
      tennessee | .0836898 .0441083 1.90 0.058 -.0027609 .1701405
      texas | .5799728 .0415557 13.96 0.000 .4985252 .6614204
      utah | -1.535726 .0559852 -27.43 0.000 -1.645455 -1.425997
      vermont | -3.110025 .074864 -41.54 0.000 -3.256755 -2.963294
      virginia | -.1371805 .0433857 -3.16 0.002 -.2222149 -.052146
      washington | -.7365939 .0457942 -16.08 0.000 -.8263489 -.6468388
      west_virginia | -1.323135 .0539866 -24.51 0.000 -1.428947 -1.217323
      wisconsin | -1.127178 .0499649 -22.56 0.000 -1.225108 -1.029249
      wyoming | -2.685405 .0677426 -39.64 0.000 -2.818178 -2.552632
      _cons | .9808652 .0376085 26.08 0.000 .907154 1.054576
      ----------------------------+----------------------------------------------------------------
      /ln_r | 7.484747 . . .
      /ln_s | 6.678558 .0110904 6.656821 6.700295
      ----------------------------+----------------------------------------------------------------
      r | 1780.674 . . .
      s | 795.1715 8.818793 778.0734 812.6452
      ---------------------------------------------------------------------------------------------
      LR test vs. pooled: chibar2(01) = 0.00 Prob >= chibar2 = 1.000
      convergence not achieved
      r(430);

      Comment


      • #4
        Please see the FAQ for how to use code blocks so that the posted output is more easily readable.

        Despite the jumble, there are two things that grab my eye.

        1. Sometimes when independent variables in a model range over a wide range of orders of magnitude, you encounter convergence difficulties. In this case, two variables, STATE_RESIDENTS and STATE_PRISONERS appear to be very large numbers (because they have extremely small regression coefficients with extremely small standard errors), whereas nearly everything else is a 0/1 dichotomy. If you rescale those two variables so that they range over smaller numbers (I would say in this case that probably means dividing each of them by 100,000 or perhaps even by 1,000,000), they will be closer in magnitude to the other variables, and you might get better results. Just remember when interpreting the results, that those coefficients are now per 100,000 or per 1,000,000 people, rather than per person.

        2. You didn't show us how you -xtset- the data before running this. From what you wrote in your original post, I (mis?)understood that state was the "panel" variable here. If so, it doesn't make sense to have indicator variables for the states in the model at the same time that you are fitting a random effect for state, and you should remove them.

        Finally, I'll comment on the r and s values near the end of the output. These are the parameters of a beta distribution that is fit to the random effects in your model (these being reciprocal of dispersion). Their values are pretty large, and a beta distribution with those values of r and s has variance approximately 0.00008. That's pretty small--pretty close to a variance component being zero, though I have seen models that successfully converged with values like this. It may be that your random effects just have so little variation that it is difficult to fit the model to them. BUT, that paucity of variation state random effects could also arise as a result of what I pointed out in #2--so look and see if the values of r and s shrink substantially if the state dummies are removed.

        Comment


        • #5
          Yes, you are correct. I used the extset commend like "xtset REF_STATE REF_TIME." Really appreciate you for your advice.

          Comment

          Working...
          X