I am using unbalanced panel data on individual respondents (T = 11 and N = 83687). I am running a conditional logit model with fixed effects, the dependent variable it's a dummy (at time t) and I run most of the covariates with a lag of one year in (t-1).
I'm using regional (i.region) and year's effects (i.syear). My problem is due to convergence. If I run the fixed effects conditional logit model with all the year's effects (i.syear) and the covariates Age and Age2 (using squared Age to model the age effect) I get no convergence. Therefore, considering that I got no standard errors (.) in the year effect 2019, I tried to create myself a dummy variable for each year I'm considering and I excluded the year 2019 effect (because I thought there were not enough observations to achieve the convergence). In that case, it works fine and the convergence it's achieved.
Moreover, I tried to consider instead of Age and Age2 some Age Categories ( 5 categories). With the Age categories everything it's working fine and the convergence it's achieved considering all the year's effects (by using i.syear).
I think I will probably consider the Age categories model for my final version but I was wondering why was the convergence not achieved when considering the Age and Age2 and all the year's effects (i.syear) and on the contrary, it was when considering the Age categories and all the year's effects?
here there is the code:
Code:
. xtlogit resigning L.insecurity L.ln_income i.region i.syear L.age L.age2 i.L.married i.L.health_status1 L.ye > ars_of_education L.n_of_children L.annual_working_hours, fe iterate(20) note: multiple positive outcomes within groups encountered. note: 12,229 groups (50,752 obs) omitted because of all positive or all negative outcomes. Iteration 0: log likelihood = -4954.8824 (not concave) Iteration 1: log likelihood = -4923.8604 (not concave) Iteration 2: log likelihood = -4918.2668 (not concave) Iteration 3: log likelihood = -4917.9681 (not concave) Iteration 4: log likelihood = -4917.8448 (not concave) Iteration 5: log likelihood = -4917.7838 (not concave) Iteration 6: log likelihood = -4917.6602 (not concave) Iteration 7: log likelihood = -4917.6032 (not concave) Iteration 8: log likelihood = -4917.3976 (not concave) Iteration 9: log likelihood = -4917.3279 (not concave) Iteration 10: log likelihood = -4917.3174 (not concave) Iteration 11: log likelihood = -4917.3072 (not concave) Iteration 12: log likelihood = -4917.2688 (not concave) Iteration 13: log likelihood = -4917.2554 (not concave) Iteration 14: log likelihood = -4917.2542 (not concave) Iteration 15: log likelihood = -4917.2531 (not concave) Iteration 16: log likelihood = -4917.252 (not concave) Iteration 17: log likelihood = -4917.2511 (not concave) Iteration 18: log likelihood = -4917.2502 (not concave) Iteration 19: log likelihood = -4917.2493 (not concave) Iteration 20: log likelihood = -4917.2485 (not concave) convergence not achieved Conditional fixed-effects logistic regression Number of obs = 13,872 Group variable: pid Number of groups = 2,474 Obs per group: min = 2 avg = 5.6 max = 10 LR chi2(34) = 196.67 Log likelihood = -4917.2485 Prob > chi2 = 0.0000 resigning Coefficient Std. err. z P>z [95% conf. interval] insecurity L1. .1586778 .0395872 4.01 0.000 .0810883 .2362673 ln_income L1. -.0959634 .1089586 -0.88 0.378 -.3095183 .1175914 region 2 -.0029459 .6049219 -0.00 0.996 -1.188571 1.182679 3 .0213726 .8443094 0.03 0.980 -1.633444 1.676189 4 54.71097 . . . . . 5 -1.05323 .7851114 -1.34 0.180 -2.59202 .4855604 6 -.3901004 .8363038 -0.47 0.641 -2.029226 1.249025 7 .6840851 1.003371 0.68 0.495 -1.282485 2.650655 8 .5695272 .8315001 0.68 0.493 -1.060183 2.199237 9 .0330805 .8980872 0.04 0.971 -1.727138 1.793299 10 -43.49452 1.72e+09 -0.00 1.000 -3.38e+09 3.38e+09 11 .9676775 .9933413 0.97 0.330 -.9792356 2.914591 12 .7399395 1.002769 0.74 0.461 -1.225452 2.705332 13 .8720612 1.059855 0.82 0.411 -1.205217 2.949339 14 -.1655239 1.044938 -0.16 0.874 -2.213564 1.882517 15 -1.174527 1.385044 -0.85 0.396 -3.889163 1.540109 16 -1.920376 1.317821 -1.46 0.145 -4.503258 .6625068 syear 2011 .1142141 .0922802 1.24 0.216 -.0666517 .2950799 2012 .162089 .0917376 1.77 0.077 -.0177134 .3418914 2013 .2067589 .0895892 2.31 0.021 .0311674 .3823505 2014 .1426668 .0795742 1.79 0.073 -.0132957 .2986292 2015 .0924091 .0750384 1.23 0.218 -.0546635 .2394818 2016 .0959823 .0722217 1.33 0.184 -.0455696 .2375341 2017 .0659704 .0703078 0.94 0.348 -.0718303 .2037711 2018 .1146478 .0704689 1.63 0.104 -.0234688 .2527643 2019 .0564252 . . . . . age L1. -.1092868 .0493051 -2.22 0.027 -.205923 -.0126507 age2 L1. .0007596 .0005432 1.40 0.162 -.0003051 .0018242 L.married 1 -.1836457 .1240138 -1.48 0.139 -.4267083 .0594169 L.health_status1 2 .4764549 .2225177 2.14 0.032 .0403282 .9125817 3 .1984931 .2217544 0.90 0.371 -.2361375 .6331237 4 .1046584 .2238471 0.47 0.640 -.3340738 .5433906 5 .2330741 .2369465 0.98 0.325 -.2313325 .6974807 years_of_education L1. -.0513078 .0984258 -0.52 0.602 -.2442188 .1416031 n_of_children L1. -.1159357 .0600321 -1.93 0.053 -.2335965 .0017251 annual_working_hours L1. -.0003312 .0000471 -7.03 0.000 -.0004236 -.0002388
Code:
. xtlogit resigning L.insecurity L.ln_income $yearsdummy i.region L.age L.age2 i.L.sex i.L.married i.L.health_ > status1 L.years_of_education L.n_of_children L.annual_working_hours, fe iterate(20) note: multiple positive outcomes within groups encountered. note: 12,229 groups (50,752 obs) omitted because of all positive or all negative outcomes. note: 1L.sex omitted because of no within-group variance. Iteration 0: log likelihood = -4953.8719 Iteration 1: log likelihood = -4917.7094 Iteration 2: log likelihood = -4917.2589 Iteration 3: log likelihood = -4917.2287 Iteration 4: log likelihood = -4917.2222 Iteration 5: log likelihood = -4917.2207 Iteration 6: log likelihood = -4917.2204 Iteration 7: log likelihood = -4917.2203 Iteration 8: log likelihood = -4917.2203 Conditional fixed-effects logistic regression Number of obs = 13,872 Group variable: pid Number of groups = 2,474 Obs per group: min = 2 avg = 5.6 max = 10 LR chi2(35) = 196.73 Log likelihood = -4917.2203 Prob > chi2 = 0.0000 resigning Coefficient Std. err. z P>z [95% conf. interval] insecurity L1. .1595486 .0395919 4.03 0.000 .08195 .2371473 ln_income L1. -.0939751 .1089648 -0.86 0.388 -.3075422 .119592 year11 .1067305 .0922694 1.16 0.247 -.0741142 .2875752 year12 .1484747 .0917286 1.62 0.106 -.03131 .3282595 year13 .1870052 .0895873 2.09 0.037 .0114174 .3625931 year14 .1159637 .0795739 1.46 0.145 -.0399984 .2719258 year15 .0599535 .075038 0.80 0.424 -.0871184 .2070253 year16 .0574416 .0722222 0.80 0.426 -.0841113 .1989945 year17 .0212388 .0703086 0.30 0.763 -.1165636 .1590412 year18 .0639982 .0704692 0.91 0.364 -.074119 .2021153 region 2 -.010283 .6049746 -0.02 0.986 -1.196011 1.175445 3 .0200304 .8440641 0.02 0.981 -1.634305 1.674366 4 23.77858 26989.53 0.00 0.999 -52874.74 52922.3 5 -1.056672 .7848169 -1.35 0.178 -2.594885 .4815409 6 -.3918029 .836058 -0.47 0.639 -2.030447 1.246841 7 .6866278 1.003021 0.68 0.494 -1.279258 2.652514 8 .5670053 .8313347 0.68 0.495 -1.062381 2.196391 9 .0289339 .8978407 0.03 0.974 -1.730801 1.788669 10 -14.35951 810.0542 -0.02 0.986 -1602.037 1573.318 11 .966391 .9924808 0.97 0.330 -.9788356 2.911618 12 .7433985 1.002162 0.74 0.458 -1.220804 2.707601 13 .87058 1.059877 0.82 0.411 -1.206741 2.947901 14 -.1643945 1.044952 -0.16 0.875 -2.212463 1.883674 15 -1.177462 1.385052 -0.85 0.395 -3.892115 1.53719 16 -1.923164 1.317747 -1.46 0.144 -4.505901 .6595729 age L1. -.1143711 .0492992 -2.32 0.020 -.2109957 -.0177464 age2 L1. .000883 .0005431 1.63 0.104 -.0001814 .0019474 L.sex 1 0 (omitted) L.married 1 -.1807483 .1240086 -1.46 0.145 -.4238007 .0623041 L.health_status1 2 .4649026 .2221624 2.09 0.036 .0294722 .9003329 3 .1864222 .2213913 0.84 0.400 -.2474967 .6203412 4 .0925622 .2234899 0.41 0.679 -.3454699 .5305943 5 .2207287 .2366089 0.93 0.351 -.2430162 .6844736 years_of_education L1. -.0487996 .0984007 -0.50 0.620 -.2416614 .1440622 n_of_children L1. -.112988 .0600387 -1.88 0.060 -.2306617 .0046858 annual_working_hours L1. -.0003305 .0000471 -7.01 0.000 -.0004229 -.0002381
Code:
. xtlogit resigning L.insecurity L.ln_income i.syear i.region i.L.age_cate i.L.sex i.L.married i.L.health_stat > us1 L.years_of_education L.n_of_children L.annual_working_hours, fe iterate(20) note: multiple positive outcomes within groups encountered. note: 12,207 groups (50,651 obs) omitted because of all positive or all negative outcomes. note: 1L.sex omitted because of no within-group variance. Iteration 0: log likelihood = -4954.8564 Iteration 1: log likelihood = -4914.5 Iteration 2: log likelihood = -4914.0655 Iteration 3: log likelihood = -4914.0341 Iteration 4: log likelihood = -4914.0284 Iteration 5: log likelihood = -4914.0271 Iteration 6: log likelihood = -4914.0269 Iteration 7: log likelihood = -4914.0268 Iteration 8: log likelihood = -4914.0268 Conditional fixed-effects logistic regression Number of obs = 13,866 Group variable: pid Number of groups = 2,474 Obs per group: min = 2 avg = 5.6 max = 10 LR chi2(37) = 201.15 Log likelihood = -4914.0268 Prob > chi2 = 0.0000 resigning Coefficient Std. err. z P>z [95% conf. interval] insecurity L1. .1568622 .0395197 3.97 0.000 .079405 .2343195 ln_income L1. -.0901518 .1090069 -0.83 0.408 -.3038015 .1234978 syear 2011 .0686782 .0977131 0.70 0.482 -.1228359 .2601924 2012 .0736324 .1028235 0.72 0.474 -.1278979 .2751627 2013 .080782 .1058594 0.76 0.445 -.1266987 .2882627 2014 -.0216384 .1040958 -0.21 0.835 -.2256624 .1823856 2015 -.1144291 .1060272 -1.08 0.280 -.3222386 .0933803 2016 -.1529156 .1085844 -1.41 0.159 -.3657372 .059906 2017 -.2264816 .1120058 -2.02 0.043 -.446009 -.0069543 2018 -.2151943 .1162956 -1.85 0.064 -.4431294 .0127408 2019 -.319139 .1217926 -2.62 0.009 -.557848 -.0804299 region 2 -.0053282 .6031277 -0.01 0.993 -1.187437 1.17678 3 -.0144088 .8435746 -0.02 0.986 -1.667785 1.638967 4 23.9593 29766.11 0.00 0.999 -58316.55 58364.47 5 -1.077494 .7862965 -1.37 0.171 -2.618607 .4636191 6 -.4071589 .8362864 -0.49 0.626 -2.04625 1.231932 7 .658063 1.002816 0.66 0.512 -1.30742 2.623546 8 .5285258 .8313712 0.64 0.525 -1.100932 2.157983 9 .0023187 .8976535 0.00 0.998 -1.75705 1.761687 10 -14.84499 1041.706 -0.01 0.989 -2056.552 2026.862 11 .9501312 .9967343 0.95 0.340 -1.003432 2.903694 12 .6816164 1.005483 0.68 0.498 -1.289094 2.652327 13 .8610184 1.057921 0.81 0.416 -1.212469 2.934506 14 -.2024885 1.044481 -0.19 0.846 -2.249633 1.844656 15 -1.178271 1.384005 -0.85 0.395 -3.89087 1.534328 16 -1.944648 1.316166 -1.48 0.140 -4.524285 .6349899 L.age_cate1 2 -.3518934 .1504093 -2.34 0.019 -.6466902 -.0570965 3 -.3433212 .1950824 -1.76 0.078 -.7256756 .0390332 4 -.2346913 .2379139 -0.99 0.324 -.700994 .2316115 L.sex 1 0 (omitted) L.married 1 -.1878148 .1241472 -1.51 0.130 -.4311388 .0555092 L.health_status1 2 .4630734 .222156 2.08 0.037 .0276556 .8984911 3 .1842551 .2213972 0.83 0.405 -.2496754 .6181855 4 .0916305 .2234936 0.41 0.682 -.346409 .5296699 5 .2184561 .2366326 0.92 0.356 -.2453354 .6822476 years_of_education L1. -.0569611 .0980777 -0.58 0.561 -.2491898 .1352676 n_of_children L1. -.1095408 .0597734 -1.83 0.067 -.2266946 .007613 annual_working_hours L1. -.0003309 .0000471 -7.02 0.000 -.0004233 -.0002385
Alessandro
0 Response to Fixed effects conditional logit model, problems with Age and Year's effect
Post a Comment