1) In 2013 (for example), the 95% CIs for M, F, and U do not overlap. Can I therefore conclude that the expected probabilities are statistically significantly different (at 5% level) for these 3 groups?
2) For the F group the CIs do not overlap for 2013 v 2008. Can I conclude the predicted probabilities are statistically significant (5% level) for those 2 years?
I understand if the CIs do overlap, they may or may not be statistically significantly different.
Code:
. mlogit gender_n year, vce(cluster person) Iteration 0: log pseudolikelihood = -10918.203 Iteration 1: log pseudolikelihood = -10906.418 Iteration 2: log pseudolikelihood = -10906.402 Iteration 3: log pseudolikelihood = -10906.402 Multinomial logistic regression Number of obs = 12,899 Wald chi2(2) = 33.46 Prob > chi2 = 0.0000 Log pseudolikelihood = -10906.402 Pseudo R2 = 0.0011 (Std. Err. adjusted for 7,014 clusters in person) ------------------------------------------------------------------------------ | Robust gender_n | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- Female | year | .048562 .0143528 3.38 0.001 .020431 .0766931 _cons | -98.68828 28.86279 -3.42 0.001 -155.2583 -42.11825 -------------+---------------------------------------------------------------- Male | (base outcome) -------------+---------------------------------------------------------------- U | year | -.0724472 .0184879 -3.92 0.000 -.1086829 -.0362116 _cons | 143.8082 37.17114 3.87 0.000 70.95414 216.6623 ------------------------------------------------------------------------------ . margins, at(year = (2008(1)2013)) Adjusted predictions Number of obs = 12,899 Model VCE : Robust 1._predict : Pr(gender_n==Female), predict(pr outcome(1)) 2._predict : Pr(gender_n==Male), predict(pr outcome(2)) 3._predict : Pr(gender_n==U), predict(pr outcome(3)) 1._at : year = 2008 2._at : year = 2009 3._at : year = 2010 4._at : year = 2011 5._at : year = 2012 6._at : year = 2013 ------------------------------------------------------------------------------ | Delta-method | Margin Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- _predict#_at | 1 1 | .2060591 .0085704 24.04 0.000 .1892615 .2228567 1 2 | .2160034 .0071416 30.25 0.000 .2020061 .2300007 1 3 | .226174 .0060819 37.19 0.000 .2142537 .2380943 1 4 | .2365643 .0058009 40.78 0.000 .2251948 .2479338 1 5 | .2471675 .0065827 37.55 0.000 .2342657 .2600693 1 6 | .2579761 .0082736 31.18 0.000 .2417602 .274192 2 1 | .6677101 .0101316 65.90 0.000 .6478526 .6875676 2 2 | .6667553 .0082352 80.96 0.000 .6506146 .6828961 2 3 | .6650561 .0069231 96.06 0.000 .6514871 .6786251 2 4 | .6626354 .0065492 101.18 0.000 .6497993 .6754715 2 5 | .6595178 .0072932 90.43 0.000 .6452235 .6738121 2 6 | .6557291 .0089327 73.41 0.000 .6382213 .6732369 3 1 | .1262308 .0073137 17.26 0.000 .1118962 .1405653 3 2 | .1172412 .0057115 20.53 0.000 .1060469 .1284355 3 3 | .1087699 .0046996 23.14 0.000 .0995588 .117981 3 4 | .1008003 .0043574 23.13 0.000 .0922599 .1093406 3 5 | .0933147 .0045779 20.38 0.000 .0843422 .1022873 3 6 | .0862948 .0051102 16.89 0.000 .0762791 .0963105 ------------------------------------------------------------------------------ . marginsplot Variables that uniquely identify margins: year _outcome
0 Response to non-overlapping CIs in predicted probabilities after mlogit: can we conclude statistically significant difference?
Post a Comment