Here are the individual odds ratio results by thing: logit causedecline hadaskid, vce(robust) or ------------------------------------------------------------------------------------------------------------------------------------------------------------- -> thing = 1 (social media) Iteration 0: log pseudolikelihood = -875.88119 Iteration 1: log pseudolikelihood = -872.83851 Iteration 2: log pseudolikelihood = -872.82446 Iteration 3: log pseudolikelihood = -872.82446 Logistic regression Number of obs = 1500 Wald chi2(1) = 6.30 Prob > chi2 = 0.0121 Log pseudolikelihood = -872.82446 Pseudo R2 = 0.0035 ------------------------------------------------------------------------------ | Robust causedecline | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- hadaskid | .6746242 .1057816 -2.51 0.012 .4961269 .9173413 _cons | 2.864458 .1826591 16.50 0.000 2.527921 3.245797 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------------------------------------------------------------------- -> thing = 2 (smart phones) Iteration 0: log pseudolikelihood = -1035.0746 Iteration 1: log pseudolikelihood = -1032.8826 Iteration 2: log pseudolikelihood = -1032.8825 Logistic regression Number of obs = 1500 Wald chi2(1) = 4.38 Prob > chi2 = 0.0365 Log pseudolikelihood = -1032.8825 Pseudo R2 = 0.0021 ------------------------------------------------------------------------------ | Robust causedecline | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- hadaskid | .7416859 .1059585 -2.09 0.036 .560552 .9813504 _cons | 1.226714 .0693094 3.62 0.000 1.098121 1.370365 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------------------------------------------------------------------- -> thing = 3 (the internet) Iteration 0: log pseudolikelihood = -1016.3673 Iteration 1: log pseudolikelihood = -1015.2704 Iteration 2: log pseudolikelihood = -1015.2703 Logistic regression Number of obs = 1500 Wald chi2(1) = 2.20 Prob > chi2 = 0.1381 Log pseudolikelihood = -1015.2703 Pseudo R2 = 0.0011 ------------------------------------------------------------------------------ | Robust causedecline | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- hadaskid | .839902 .0988099 -1.48 0.138 .666944 1.057713 _cons | 1.496583 .092286 6.54 0.000 1.326209 1.688845 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------------------------------------------------------------------- -> thing = 4 (the radio) Iteration 0: log pseudolikelihood = -309.44018 Iteration 1: log pseudolikelihood = -308.99521 Iteration 2: log pseudolikelihood = -308.9919 Iteration 3: log pseudolikelihood = -308.9919 Logistic regression Number of obs = 1500 Wald chi2(1) = 0.93 Prob > chi2 = 0.3345 Log pseudolikelihood = -308.9919 Pseudo R2 = 0.0014 ------------------------------------------------------------------------------ | Robust causedecline | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- hadaskid | .7728686 .2063351 -0.97 0.335 .4579919 1.304228 _cons | .0677966 .0156705 -11.64 0.000 .0430985 .1066484 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------------------------------------------------------------------- -> thing = 5 (television) Iteration 0: log pseudolikelihood = -928.651 Iteration 1: log pseudolikelihood = -928.60914 Iteration 2: log pseudolikelihood = -928.60913 Logistic regression Number of obs = 1500 Wald chi2(1) = 0.08 Prob > chi2 = 0.7719 Log pseudolikelihood = -928.60913 Pseudo R2 = 0.0000 ------------------------------------------------------------------------------ | Robust causedecline | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- hadaskid | .9525909 .1595968 -0.29 0.772 .6859562 1.322868 _cons | .46875 .0733642 -4.84 0.000 .3449208 .6370348 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------------------------------------------------------------------- -> thing = 6 (reading novels) Iteration 0: log pseudolikelihood = -154.77572 Iteration 1: log pseudolikelihood = -153.45372 Iteration 2: log pseudolikelihood = -153.4117 Iteration 3: log pseudolikelihood = -153.41166 Iteration 4: log pseudolikelihood = -153.41166 Logistic regression Number of obs = 1500 Wald chi2(1) = 2.80 Prob > chi2 = 0.0945 Log pseudolikelihood = -153.41166 Pseudo R2 = 0.0088 ------------------------------------------------------------------------------ | Robust causedecline | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- hadaskid | .5489048 .1968991 -1.67 0.094 .2717441 1.108751 _cons | .0313152 .0082139 -13.21 0.000 .0187278 .0523629 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------------------------------------------------------------------- -> thing = 7 (driving cars) Iteration 0: log pseudolikelihood = -270.68156 Iteration 1: log pseudolikelihood = -268.12608 Iteration 2: log pseudolikelihood = -268.06365 Iteration 3: log pseudolikelihood = -268.06361 Iteration 4: log pseudolikelihood = -268.06361 Logistic regression Number of obs = 1500 Wald chi2(1) = 5.32 Prob > chi2 = 0.0210 Log pseudolikelihood = -268.06361 Pseudo R2 = 0.0097 ------------------------------------------------------------------------------ | Robust causedecline | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- hadaskid | .5583777 .1410303 -2.31 0.021 .3403601 .9160462 _cons | .0647773 .0118201 -15.00 0.000 .0453003 .0926285 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------------------------------------------------------------------- -> thing = 8 (single parent homes) Iteration 0: log pseudolikelihood = -1039.1327 Iteration 1: log pseudolikelihood = -1032.8142 Iteration 2: log pseudolikelihood = -1032.8136 Iteration 3: log pseudolikelihood = -1032.8136 Logistic regression Number of obs = 1500 Wald chi2(1) = 12.51 Prob > chi2 = 0.0004 Log pseudolikelihood = -1032.8136 Pseudo R2 = 0.0061 ------------------------------------------------------------------------------ | Robust causedecline | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- hadaskid | .6642993 .0768332 -3.54 0.000 .5295578 .8333247 _cons | 1.061538 .0648943 0.98 0.329 .9416723 1.196662 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------------------------------------------------------------------- -> thing = 9 (video games) Iteration 0: log pseudolikelihood = -1036.8977 Iteration 1: log pseudolikelihood = -1016.6954 Iteration 2: log pseudolikelihood = -1016.6828 Iteration 3: log pseudolikelihood = -1016.6828 Logistic regression Number of obs = 1500 Wald chi2(1) = 39.41 Prob > chi2 = 0.0000 Log pseudolikelihood = -1016.6828 Pseudo R2 = 0.0195 ------------------------------------------------------------------------------ | Robust causedecline | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- hadaskid | .4956112 .0554185 -6.28 0.000 .3980715 .6170511 _cons | 1.12095 .0716824 1.79 0.074 .9889032 1.270629 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------------------------------------------------------------------- -> thing = 10 (heavy metal) Iteration 0: log pseudolikelihood = -630.58395 Iteration 1: log pseudolikelihood = -623.90075 Iteration 2: log pseudolikelihood = -623.81572 Iteration 3: log pseudolikelihood = -623.81571 Logistic regression Number of obs = 1500 Wald chi2(1) = 12.52 Prob > chi2 = 0.0004 Log pseudolikelihood = -623.81571 Pseudo R2 = 0.0107 ------------------------------------------------------------------------------ | Robust causedecline | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- hadaskid | .5459943 .0933952 -3.54 0.000 .3904681 .7634676 _cons | .2071856 .0173128 -18.84 0.000 .1758861 .244055 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------------------------------------------------------------------- -> thing = 11 (24-hour news) Iteration 0: log pseudolikelihood = -554.35976 Iteration 1: log pseudolikelihood = -553.86066 Iteration 2: log pseudolikelihood = -553.8584 Iteration 3: log pseudolikelihood = -553.8584 Logistic regression Number of obs = 1500 Wald chi2(1) = 1.04 Prob > chi2 = 0.3083 Log pseudolikelihood = -553.8584 Pseudo R2 = 0.0009 ------------------------------------------------------------------------------ | Robust causedecline | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- hadaskid | 1.241669 .2638143 1.02 0.308 .8187533 1.883035 _cons | .1335102 .0115713 -23.23 0.000 .1126525 .1582297 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------------------------------------------------------------------- -> thing = 12 (dance clubs) Iteration 0: log pseudolikelihood = -359.4292 Iteration 1: log pseudolikelihood = -359.37061 Iteration 2: log pseudolikelihood = -359.37058 Logistic regression Number of obs = 1500 Wald chi2(1) = 0.12 Prob > chi2 = 0.7307 Log pseudolikelihood = -359.37058 Pseudo R2 = 0.0002 ------------------------------------------------------------------------------ | Robust causedecline | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- hadaskid | 1.085145 .2576201 0.34 0.731 .681408 1.728097 _cons | .0676835 .0083027 -21.95 0.000 .053219 .0860794 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------------------------------------------------------------------- -> thing = 13 (nicotine vaping) Iteration 0: log pseudolikelihood = -1012.6725 Iteration 1: log pseudolikelihood = -1011.6167 Iteration 2: log pseudolikelihood = -1011.616 Iteration 3: log pseudolikelihood = -1011.616 Logistic regression Number of obs = 1500 Wald chi2(1) = 2.12 Prob > chi2 = 0.1453 Log pseudolikelihood = -1011.616 Pseudo R2 = 0.0010 ------------------------------------------------------------------------------ | Robust causedecline | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- hadaskid | 1.554779 .4712104 1.46 0.145 .8584062 2.816075 _cons | .6724138 .0359646 -7.42 0.000 .6054934 .7467304 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------------------------------------------------------------------- -> thing = 14 (netflix) Iteration 0: log pseudolikelihood = -611.06476 Iteration 1: log pseudolikelihood = -604.53191 Iteration 2: log pseudolikelihood = -604.17203 Iteration 3: log pseudolikelihood = -604.17161 Iteration 4: log pseudolikelihood = -604.17161 Logistic regression Number of obs = 1500 Wald chi2(1) = 15.04 Prob > chi2 = 0.0001 Log pseudolikelihood = -604.17161 Pseudo R2 = 0.0113 ------------------------------------------------------------------------------ | Robust causedecline | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- hadaskid | 2.071021 .3888115 3.88 0.000 1.433444 2.992184 _cons | .1461268 .0121461 -23.14 0.000 .1241588 .1719816 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------------------------------------------------------------------- -> thing = 15 (jazz music) Iteration 0: log pseudolikelihood = -118.91046 Iteration 1: log pseudolikelihood = -117.76582 Iteration 2: log pseudolikelihood = -117.70592 Iteration 3: log pseudolikelihood = -117.70579 Iteration 4: log pseudolikelihood = -117.70579 Logistic regression Number of obs = 1500 Wald chi2(1) = 2.52 Prob > chi2 = 0.1126 Log pseudolikelihood = -117.70579 Pseudo R2 = 0.0101 ------------------------------------------------------------------------------ | Robust causedecline | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- hadaskid | 1.961908 .8332677 1.59 0.113 .8534022 4.510281 _cons | .0122526 .0034202 -15.77 0.000 .0070897 .0211753 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------------------------------------------------------------------- -> thing = 17 (long hair) Iteration 0: log pseudolikelihood = -222.57414 Iteration 1: log pseudolikelihood = -221.95506 Iteration 2: log pseudolikelihood = -221.9514 Iteration 3: log pseudolikelihood = -221.9514 Logistic regression Number of obs = 1500 Wald chi2(1) = 1.25 Prob > chi2 = 0.2641 Log pseudolikelihood = -221.9514 Pseudo R2 = 0.0028 ------------------------------------------------------------------------------ | Robust causedecline | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- hadaskid | .7272727 .2074072 -1.12 0.264 .4158603 1.271883 _cons | .0416667 .0083428 -15.87 0.000 .028142 .0616911 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------------------------------------------------------------------- -> thing = 18 (ballroom dancing) Iteration 0: log pseudolikelihood = -106.21583 Iteration 1: log pseudolikelihood = -105.82001 Iteration 2: log pseudolikelihood = -105.8017 Iteration 3: log pseudolikelihood = -105.80167 Iteration 4: log pseudolikelihood = -105.80167 Logistic regression Number of obs = 1500 Wald chi2(1) = 0.90 Prob > chi2 = 0.3426 Log pseudolikelihood = -105.80167 Pseudo R2 = 0.0039 ------------------------------------------------------------------------------ | Robust causedecline | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- hadaskid | 1.639999 .8547857 0.95 0.343 .5904564 4.555114 _cons | .0121951 .003169 -16.96 0.000 .0073282 .0202944 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------------------------------------------------------------------- -> thing = 19 (motion pictures) Iteration 0: log pseudolikelihood = -611.06476 Iteration 1: log pseudolikelihood = -606.07612 Iteration 2: log pseudolikelihood = -606.02733 Iteration 3: log pseudolikelihood = -606.02733 Logistic regression Number of obs = 1500 Wald chi2(1) = 9.28 Prob > chi2 = 0.0023 Log pseudolikelihood = -606.02733 Pseudo R2 = 0.0082 ------------------------------------------------------------------------------ | Robust causedecline | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- hadaskid | 1.744974 .3189179 3.05 0.002 1.219608 2.49665 _cons | .1082474 .0175896 -13.68 0.000 .0787231 .1488446 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------------------------------------------------------------------- -> thing = 20 (not going to church) Iteration 0: log pseudolikelihood = -991.0589 Iteration 1: log pseudolikelihood = -975.09368 Iteration 2: log pseudolikelihood = -975.02292 Iteration 3: log pseudolikelihood = -975.02292 Logistic regression Number of obs = 1500 Wald chi2(1) = 30.43 Prob > chi2 = 0.0000 Log pseudolikelihood = -975.02292 Pseudo R2 = 0.0162 ------------------------------------------------------------------------------ | Robust causedecline | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- hadaskid | .4991929 .0628753 -5.52 0.000 .3899925 .6389702 _cons | .7163462 .0444034 -5.38 0.000 .6343959 .8088826 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------------------------------------------------------------------- -> thing = 21 (online dating) Iteration 0: log pseudolikelihood = -792.88552 Iteration 1: log pseudolikelihood = -790.61786 Iteration 2: log pseudolikelihood = -790.58947 Iteration 3: log pseudolikelihood = -790.58947 Logistic regression Number of obs = 1500 Wald chi2(1) = 4.84 Prob > chi2 = 0.0278 Log pseudolikelihood = -790.58947 Pseudo R2 = 0.0029 ------------------------------------------------------------------------------ | Robust causedecline | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- hadaskid | 1.578006 .3272132 2.20 0.028 1.051004 2.369259 _cons | .2726433 .0179136 -19.78 0.000 .2397 .3101141 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------------------------------------------------------------------- -> thing = 22 (calculators) Iteration 0: log pseudolikelihood = -198.62562 Iteration 1: log pseudolikelihood = -194.95982 Iteration 2: log pseudolikelihood = -194.75722 Iteration 3: log pseudolikelihood = -194.75679 Iteration 4: log pseudolikelihood = -194.75679 Logistic regression Number of obs = 1500 Wald chi2(1) = 7.66 Prob > chi2 = 0.0056 Log pseudolikelihood = -194.75679 Pseudo R2 = 0.0195 ------------------------------------------------------------------------------ | Robust causedecline | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- hadaskid | .4247537 .1313868 -2.77 0.006 .2316539 .778816 _cons | .0478927 .0098085 -14.84 0.000 .0320584 .0715479 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------------------------------------------------------------------- -> thing = 23 (autocorrect) Iteration 0: log pseudolikelihood = -400.83165 Iteration 1: log pseudolikelihood = -400.72868 Iteration 2: log pseudolikelihood = -400.72841 Iteration 3: log pseudolikelihood = -400.72841 Logistic regression Number of obs = 1500 Wald chi2(1) = 0.20 Prob > chi2 = 0.6552 Log pseudolikelihood = -400.72841 Pseudo R2 = 0.0003 ------------------------------------------------------------------------------ | Robust causedecline | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- hadaskid | .8634454 .2838748 -0.45 0.655 .4532997 1.644691 _cons | .0827251 .0085259 -24.18 0.000 .0675942 .1012429 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------------------------------------------------------------------- -> thing = 24 (word processors) Iteration 0: log pseudolikelihood = -191.57957 Iteration 1: log pseudolikelihood = -191.19397 Iteration 2: log pseudolikelihood = -191.19148 Iteration 3: log pseudolikelihood = -191.19148 Logistic regression Number of obs = 1500 Wald chi2(1) = 0.73 Prob > chi2 = 0.3938 Log pseudolikelihood = -191.19148 Pseudo R2 = 0.0020 ------------------------------------------------------------------------------ | Robust causedecline | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- hadaskid | .7123822 .2833384 -0.85 0.394 .3267093 1.553333 _cons | .0310219 .0054039 -19.94 0.000 .0220492 .043646 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------------------------------------------------------------------- -> thing = 25 (audio/electronic books) Iteration 0: log pseudolikelihood = -264.49241 Iteration 1: log pseudolikelihood = -262.22545 Iteration 2: log pseudolikelihood = -261.621 Iteration 3: log pseudolikelihood = -261.62098 Iteration 4: log pseudolikelihood = -261.62098 Logistic regression Number of obs = 1500 Wald chi2(1) = 6.75 Prob > chi2 = 0.0094 Log pseudolikelihood = -261.62098 Pseudo R2 = 0.0109 ------------------------------------------------------------------------------ | Robust causedecline | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- hadaskid | 2.378558 .7934418 2.60 0.009 1.236999 4.573603 _cons | .039725 .0056191 -22.81 0.000 .0301066 .0524163 ------------------------------------------------------------------------------ The results from the metan or baseratese, effect(Odds ratio) Study | ES [95% Conf. Interval] % Weight ---------------------+--------------------------------------------------- 1 | 0.675 0.648 0.701 24.51 2 | 0.742 0.705 0.778 13.40 3 | 0.840 0.807 0.873 15.93 4 | 0.773 0.401 1.145 0.13 5 | 0.953 0.889 1.016 4.43 6 | 0.549 -0.370 1.468 0.02 7 | 0.558 0.113 1.004 0.09 8 | 0.664 0.624 0.705 10.88 9 | 0.496 0.454 0.537 10.15 10 | 0.546 0.414 0.678 1.02 11 | 1.242 1.080 1.403 0.68 12 | 1.085 0.782 1.388 0.19 13 | 1.555 1.506 1.603 7.57 14 | 2.071 1.932 2.210 0.92 15 | 1.962 0.684 3.240 0.01 16 | 0.727 0.151 1.304 0.05 17 | 1.640 0.170 3.110 0.01 18 | 1.745 1.606 1.884 0.92 19 | 0.499 0.447 0.552 6.42 20 | 1.578 1.489 1.667 2.26 21 | 0.425 -0.243 1.093 0.04 22 | 0.863 0.603 1.124 0.26 23 | 0.712 0.012 1.412 0.04 24 | 2.379 1.919 2.838 0.08 ---------------------+--------------------------------------------------- I-V pooled ES | 0.807 0.794 0.820 100.00 ---------------------+--------------------------------------------------- Heterogeneity chi-squared = 2326.82 (d.f. = 23) p = 0.000 I-squared (variation in ES attributable to heterogeneity) = 99.0% Test of ES=0 : z= 118.87 p = 0.000 robumeta, however, can’t work as it’s all 1 study. So ignoring this dependence, we’ve got a positive effect in the predicted direction. Meaning, the more likely someone was to have the thing growing up, the less likely they were to believe it is causing a decline in the youth Trying to get the ‘if I didn’t have it growing up it’s bad’ analysis squared away. The robumeta (robust meta-analysis) wouldn’t work, so I’m trying to take into account the nested nature of the data. The best way I can think is to run a random-effects model. Here are those results xtset id thing panel variable: id (strongly balanced) time variable: thing, 1 to 25, but with gaps delta: 1 unit . xtmixed causedecline hadaskid || id: , vce(robust) Performing EM optimization: Performing gradient-based optimization: Iteration 0: log pseudolikelihood = -18163.722 Iteration 1: log pseudolikelihood = -18163.722 Computing standard errors: Mixed-effects regression Number of obs = 36000 Group variable: id Number of groups = 1500 Obs per group: min = 24 avg = 24.0 max = 24 Wald chi2(1) = 292.45 Log pseudolikelihood = -18163.722 Prob > chi2 = 0.0000 (Std. Err. adjusted for 1500 clusters in id) ------------------------------------------------------------------------------ | Robust causedecline | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- hadaskid | -.0971822 .0056828 -17.10 0.000 -.1083202 -.0860442 _cons | .2457524 .004176 58.85 0.000 .2375676 .2539373 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------ | Robust Random-effects Parameters | Estimate Std. Err. [95% Conf. Interval] -----------------------------+------------------------------------------------ id: Identity | sd(_cons) | .1091791 .0046562 .1004241 .1186974 -----------------------------+------------------------------------------------ sd(Residual) | .3920798 .0022003 .3877909 .3964161 ------------------------------------------------------------------------------ Including a random effect for ‘thing’ doesn’t work, however. (non-convergence) .This is the same when r.thing is included as a random effect. xtmixed causedecline hadaskid i.thing || id:, vce(robust) Performing EM optimization: Performing gradient-based optimization: Iteration 0: log pseudolikelihood = -12110.378 Iteration 1: log pseudolikelihood = -12110.378 Computing standard errors: Mixed-effects regression Number of obs = 36000 Group variable: id Number of groups = 1500 Obs per group: min = 24 avg = 24.0 max = 24 Wald chi2(24) = 5654.18 Log pseudolikelihood = -12110.378 Prob > chi2 = 0.0000 (Std. Err. adjusted for 1500 clusters in id) ------------------------------------------------------------------------------ | Robust causedecline | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- hadaskid | -.0397026 .0053478 -7.42 0.000 -.050184 -.0292211 | thing | 2 | -.1895765 .0148409 -12.77 0.000 -.2186641 -.1604889 3 | -.1363837 .014853 -9.18 0.000 -.1654952 -.1072723 4 | -.6510453 .0131427 -49.54 0.000 -.6768045 -.6252861 5 | -.3903505 .0153269 -25.47 0.000 -.4203907 -.3603102 6 | -.6871165 .0125606 -54.70 0.000 -.7117349 -.662498 7 | -.6652968 .0127278 -52.27 0.000 -.6902427 -.6403508 8 | -.2377485 .0166729 -14.26 0.000 -.2704268 -.2050703 9 | -.252033 .0152801 -16.49 0.000 -.2819814 -.2220846 10 | -.5733879 .0144602 -39.65 0.000 -.6017293 -.5450465 11 | -.6079735 .0138911 -43.77 0.000 -.6351996 -.5807475 12 | -.6603523 .0129416 -51.03 0.000 -.6857174 -.6349873 13 | -.3285526 .0157558 -20.85 0.000 -.3594333 -.2976718 14 | -.5885029 .014115 -41.69 0.000 -.6161678 -.560838 15 | -.7084681 .0121829 -58.15 0.000 -.7323461 -.6845901 17 | -.6785789 .0125693 -53.99 0.000 -.7032142 -.6539435 18 | -.7149942 .0121768 -58.72 0.000 -.7388603 -.6911281 19 | -.5654225 .0143117 -39.51 0.000 -.5934728 -.5373721 20 | -.3503887 .0159559 -21.96 0.000 -.3816617 -.3191157 21 | -.510488 .0146286 -34.90 0.000 -.5391595 -.4818166 22 | -.6805193 .0126446 -53.82 0.000 -.7053023 -.6557362 23 | -.6553764 .0130751 -50.12 0.000 -.6810032 -.6297496 24 | -.6972837 .0120924 -57.66 0.000 -.7209843 -.673583 25 | -.6887312 .012443 -55.35 0.000 -.713119 -.6643434 | _cons | .735077 .0114801 64.03 0.000 .7125765 .7575775 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------ | Robust Random-effects Parameters | Estimate Std. Err. [95% Conf. Interval] -----------------------------+------------------------------------------------ id: Identity | sd(_cons) | .1158445 .0041955 .1079065 .1243665 -----------------------------+------------------------------------------------ sd(Residual) | .3291375 .0018224 .325585 .3327288 ------------------------------------------------------------------------------ Conditioning on order effects xtmixed causedecline hadaskid causes1st i.thing || id:, vce(robust) Performing EM optimization: Performing gradient-based optimization: Iteration 0: log pseudolikelihood = -12106.559 Iteration 1: log pseudolikelihood = -12106.559 Computing standard errors: Mixed-effects regression Number of obs = 36000 Group variable: id Number of groups = 1500 Obs per group: min = 24 avg = 24.0 max = 24 Wald chi2(25) = 5654.01 Log pseudolikelihood = -12106.559 Prob > chi2 = 0.0000 (Std. Err. adjusted for 1500 clusters in id) ------------------------------------------------------------------------------ | Robust causedecline | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- hadaskid | -.0400372 .0053572 -7.47 0.000 -.0505372 -.0295372 causes1st | -.0190965 .0069122 -2.76 0.006 -.0326442 -.0055489 | thing | 2 | -.1895729 .014841 -12.77 0.000 -.2186607 -.1604852 3 | -.136342 .0148535 -9.18 0.000 -.1654543 -.1072297 4 | -.6508293 .0131406 -49.53 0.000 -.6765843 -.6250743 5 | -.3901062 .0153263 -25.45 0.000 -.4201452 -.3600671 6 | -.6869404 .0125576 -54.70 0.000 -.7115529 -.6623279 7 | -.6651279 .0127272 -52.26 0.000 -.6900727 -.640183 8 | -.2377014 .016673 -14.26 0.000 -.2703799 -.205023 9 | -.2519659 .0152806 -16.49 0.000 -.2819154 -.2220163 10 | -.5733265 .0144599 -39.65 0.000 -.6016674 -.5449856 11 | -.6079733 .0138909 -43.77 0.000 -.635199 -.5807476 12 | -.660316 .0129412 -51.02 0.000 -.6856802 -.6349517 13 | -.3285909 .0157562 -20.85 0.000 -.3594726 -.2977093 14 | -.5885071 .0141151 -41.69 0.000 -.6161722 -.5608421 15 | -.7084215 .0121821 -58.15 0.000 -.7322979 -.6845451 17 | -.6784376 .0125679 -53.98 0.000 -.7030703 -.653805 18 | -.7149857 .0121764 -58.72 0.000 -.7388511 -.6911203 19 | -.5652322 .0143121 -39.49 0.000 -.5932833 -.537181 20 | -.3503414 .0159557 -21.96 0.000 -.381614 -.3190688 21 | -.510509 .0146285 -34.90 0.000 -.5391803 -.4818377 22 | -.6803551 .0126429 -53.81 0.000 -.7051348 -.6555754 23 | -.655388 .013075 -50.13 0.000 -.6810145 -.6297614 24 | -.6972495 .0120919 -57.66 0.000 -.7209492 -.6735499 25 | -.6887486 .0124432 -55.35 0.000 -.7131367 -.6643605 | _cons | .7445845 .0120769 61.65 0.000 .7209142 .7682548 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------ | Robust Random-effects Parameters | Estimate Std. Err. [95% Conf. Interval] -----------------------------+------------------------------------------------ id: Identity | sd(_cons) | .1154577 .004216 .1074832 .1240238 -----------------------------+------------------------------------------------ sd(Residual) | .3291369 .0018224 .3255844 .3327281 ------------------------------------------------------------------------------ Results don't change when controlling for age xtmixed causedecline hadaskid age i.thing || id:, vce(robust) Performing EM optimization: Performing gradient-based optimization: Iteration 0: log pseudolikelihood = -12101.079 Iteration 1: log pseudolikelihood = -12101.079 Computing standard errors: Mixed-effects regression Number of obs = 36000 Group variable: id Number of groups = 1500 Obs per group: min = 24 avg = 24.0 max = 24 Wald chi2(25) = 5657.66 Log pseudolikelihood = -12101.079 Prob > chi2 = 0.0000 (Std. Err. adjusted for 1500 clusters in id) ------------------------------------------------------------------------------ | Robust causedecline | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- hadaskid | -.0390497 .0053185 -7.34 0.000 -.0494737 -.0286257 age | .0008972 .0002079 4.32 0.000 .0004898 .0013046 | thing | 2 | -.1895835 .0148407 -12.77 0.000 -.2186708 -.1604962 3 | -.1364651 .0148529 -9.19 0.000 -.1655763 -.107354 4 | -.6514666 .0131434 -49.57 0.000 -.6772271 -.6257061 5 | -.390827 .0153189 -25.51 0.000 -.4208516 -.3608025 6 | -.6874598 .0125618 -54.73 0.000 -.7120805 -.6628392 7 | -.6656262 .0127274 -52.30 0.000 -.6905714 -.640681 8 | -.2378403 .0166727 -14.27 0.000 -.2705182 -.2051624 9 | -.252164 .0152796 -16.50 0.000 -.2821114 -.2222166 10 | -.5735076 .014458 -39.67 0.000 -.6018447 -.5451704 11 | -.607974 .0138914 -43.77 0.000 -.6352007 -.5807473 12 | -.6604233 .0129424 -51.03 0.000 -.6857899 -.6350566 13 | -.3284777 .0157571 -20.85 0.000 -.3593611 -.2975943 14 | -.5884946 .0141147 -41.69 0.000 -.6161589 -.5608304 15 | -.7085591 .0121838 -58.16 0.000 -.7324388 -.6846793 17 | -.6788544 .012568 -54.01 0.000 -.7034873 -.6542214 18 | -.7150107 .0121772 -58.72 0.000 -.7388776 -.6911438 19 | -.5657937 .0142984 -39.57 0.000 -.5938181 -.5377694 20 | -.350481 .0159552 -21.97 0.000 -.3817525 -.3192094 21 | -.5104471 .0146286 -34.89 0.000 -.5391186 -.4817756 22 | -.6808396 .0126437 -53.85 0.000 -.7056208 -.6560584 23 | -.6553537 .0130751 -50.12 0.000 -.6809804 -.629727 24 | -.6973503 .0120934 -57.66 0.000 -.7210528 -.6736477 25 | -.6886973 .0124429 -55.35 0.000 -.7130848 -.6643097 | _cons | .6900617 .0155614 44.34 0.000 .6595619 .7205615 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------ | Robust Random-effects Parameters | Estimate Std. Err. [95% Conf. Interval] -----------------------------+------------------------------------------------ id: Identity | sd(_cons) | .1148716 .0042594 .1068194 .1235308 -----------------------------+------------------------------------------------ sd(Residual) | .3291389 .0018224 .3255864 .3327301 ------------------------------------------------------------------------------