Announcement

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

  • gologit2 v mlogit

    I have an ordinal dependent variable, a grouping of the number of joints with pain. If I fit a a gologit2 model with option pl, I get identical output to an ologit model. However, with an unconstrained model, if I fit a gologit2 model with option npl, I do not get the same results as an mlogit model. Why?

    My interest in this model is in the dep(ression) term. Note that in the gologit2 model it appears protective, and in the mlogit model it is associated with increased risk (as it should be).

    Graeme Tucker

    . gologit2 jointcat age bmi dep i.edn sf1_ asthma i.pa sex diabetes smoker anxiety i.osteo i.income pension [pw=wt] , vce(cluster id) or

    Generalized Ordered Logit Estimates Number of obs = 2,727
    Wald chi2(69) = 472.50
    Prob > chi2 = 0.0000
    Log pseudolikelihood = -3328.3 Pseudo R2 = 0.1009

    (Std. Err. adjusted for 2,036 clusters in id)
    ----------------------------------------------------------------------------------------------------------------
    | Robust
    jointcat | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval]
    -----------------------------------------------+----------------------------------------------------------------
    4-6 |
    age | .9794932 .0072446 -2.80 0.005 .9653964 .9937959
    bmi | .9666104 .0145221 -2.26 0.024 .9385627 .9954964
    dep | .3901502 .0748663 -4.90 0.000 .2678514 .5682895
    |
    edn |
    Trade/apprenticeship | 1.054599 .3079988 0.18 0.856 .5949652 1.869319
    Certificate/diploma | .9181697 .1988541 -0.39 0.693 .6005824 1.403697
    Bachelor degree or higher | 1.619228 .3655151 2.14 0.033 1.04031 2.520306
    |
    sf1_ | .379819 .0690861 -5.32 0.000 .2659193 .5425049
    asthma | .5589972 .123899 -2.62 0.009 .3620304 .8631261
    |
    pa |
    Active but not sufficiently active for health | 1.032043 .2002906 0.16 0.871 .7055102 1.509705
    Sufficiently active for health | .7180748 .1523106 -1.56 0.118 .473829 1.088223
    |
    sex | .8469321 .1479436 -0.95 0.342 .6013942 1.192719
    diabetes | 1.163357 .2371364 0.74 0.458 .7801992 1.734683
    smoker | .7802446 .1217418 -1.59 0.112 .5746687 1.059361
    anxiety | .7782564 .2113233 -0.92 0.356 .4570801 1.325113
    |
    osteo |
    Osteoporosis | .4168925 .1210954 -3.01 0.003 .2359252 .7366715
    Don't know/ not stated | 1.155519 .654133 0.26 0.798 .3809924 3.504593
    |
    income |
    $12,001 to $20,000 | 1.576793 .6798163 1.06 0.291 .6773125 3.670798
    $20,001 to $40,000 | 1.384754 .5521583 0.82 0.414 .6338162 3.025392
    $40,001 to $60,000 | 1.533337 .6672004 0.98 0.326 .6535137 3.597664
    $60,001 to $80,000 | 1.549335 .6997618 0.97 0.332 .6392899 3.754854
    More than $80,000 | 3.034188 1.418101 2.37 0.018 1.213984 7.58354
    Not stated | 1.026652 .4714337 0.06 0.954 .4174034 2.525172
    |
    pension | 1.177217 .2558259 0.75 0.453 .7689147 1.802331
    _cons | 65.664 54.5371 5.04 0.000 12.89329 334.4191
    -----------------------------------------------+----------------------------------------------------------------
    2-3 |
    age | .9861962 .0058217 -2.35 0.019 .9748516 .9976729
    bmi | .9566471 .0106593 -3.98 0.000 .9359817 .9777687
    dep | .6819255 .1131742 -2.31 0.021 .4925714 .9440711
    |
    edn |
    Trade/apprenticeship | 1.425037 .2951698 1.71 0.087 .9495448 2.138634
    Certificate/diploma | .7531101 .1162878 -1.84 0.066 .5564477 1.019278
    Bachelor degree or higher | 1.822836 .3378584 3.24 0.001 1.267592 2.621294
    |
    sf1_ | .2853915 .0630546 -5.68 0.000 .1850869 .4400542
    asthma | .7535789 .1304938 -1.63 0.102 .5366981 1.058102
    |
    pa |
    Active but not sufficiently active for health | 1.037002 .1768659 0.21 0.831 .74234 1.448626
    Sufficiently active for health | .6993341 .1121068 -2.23 0.026 .5107788 .9574952
    |
    sex | .8430427 .1086312 -1.33 0.185 .654888 1.085256
    diabetes | 1.00751 .1687732 0.04 0.964 .7255384 1.399067
    smoker | .8374958 .1056003 -1.41 0.160 .6541151 1.072287
    anxiety | .9242658 .2212894 -0.33 0.742 .578097 1.477723
    |
    osteo |
    Osteoporosis | .5376649 .1174422 -2.84 0.004 .3504158 .8249729
    Don't know/ not stated | .5866788 .268959 -1.16 0.245 .2388761 1.440881
    |
    income |
    $12,001 to $20,000 | 1.772012 .7006991 1.45 0.148 .8163563 3.846393
    $20,001 to $40,000 | 1.449292 .5429635 0.99 0.322 .6954427 3.020303
    $40,001 to $60,000 | 1.341675 .532918 0.74 0.459 .6159523 2.922453
    $60,001 to $80,000 | 1.317693 .546643 0.67 0.506 .5843792 2.971214
    More than $80,000 | 1.395132 .5576951 0.83 0.405 .6373133 3.054061
    Not stated | .7058026 .301059 -0.82 0.414 .3059176 1.628403
    |
    pension | 1.00358 .1652565 0.02 0.983 .7267533 1.385852
    _cons | 10.61056 7.70141 3.25 0.001 2.558086 44.01105
    -----------------------------------------------+----------------------------------------------------------------
    1 |
    age | .9795931 .0068913 -2.93 0.003 .9661792 .9931933
    bmi | .9617368 .0132121 -2.84 0.005 .9361871 .9879838
    dep | .4472497 .1148906 -3.13 0.002 .2703281 .739961
    |
    edn |
    Trade/apprenticeship | 1.379135 .3844781 1.15 0.249 .7985588 2.381806
    Certificate/diploma | 1.00379 .1877769 0.02 0.984 .6956811 1.448356
    Bachelor degree or higher | 2.139797 .4323633 3.76 0.000 1.44006 3.179543
    |
    sf1_ | .2757956 .0846773 -4.20 0.000 .1510928 .5034207
    asthma | .6429028 .1423962 -1.99 0.046 .4164983 .9923787
    |
    pa |
    Active but not sufficiently active for health | .8771243 .191287 -0.60 0.548 .5720415 1.344915
    Sufficiently active for health | .8053993 .1659279 -1.05 0.294 .5378327 1.206078
    |
    sex | .8666268 .1344095 -0.92 0.356 .6394624 1.17449
    diabetes | 1.077818 .236048 0.34 0.732 .701661 1.655629
    smoker | .9058696 .1344346 -0.67 0.505 .6772423 1.211678
    anxiety | .8957074 .3295407 -0.30 0.765 .4355115 1.842183
    |
    osteo |
    Osteoporosis | .6762433 .1775105 -1.49 0.136 .4042657 1.131199
    Don't know/ not stated | .3721351 .2103714 -1.75 0.080 .1228878 1.126919
    |
    income |
    $12,001 to $20,000 | 1.027841 .5238961 0.05 0.957 .3784986 2.791181
    $20,001 to $40,000 | 1.039662 .4725029 0.09 0.932 .4266191 2.533635
    $40,001 to $60,000 | .9052672 .4345047 -0.21 0.836 .3533639 2.319164
    $60,001 to $80,000 | 1.034682 .5195088 0.07 0.946 .3867435 2.768159
    More than $80,000 | .890918 .4264127 -0.24 0.809 .3486858 2.276362
    Not stated | .8089087 .4093885 -0.42 0.675 .2999899 2.181184
    |
    pension | 1.177691 .2653922 0.73 0.468 .757205 1.831679
    _cons | 4.724545 3.761565 1.95 0.051 .9923266 22.49393
    ----------------------------------------------------------------------------------------------------------------
    Note: _cons estimates baseline odds.

    . mlogit jointcat age bmi dep i.edn sf1_ asthma i.pa sex diabetes smoker anxiety i.osteo i.income pension [pw=wt] , vce(cluster id) rrr baseoutcome(4)

    Iteration 0: log pseudolikelihood = -3766.7674
    Iteration 1: log pseudolikelihood = -3416.51
    Iteration 2: log pseudolikelihood = -3379.5241
    Iteration 3: log pseudolikelihood = -3378.7169
    Iteration 4: log pseudolikelihood = -3378.7162
    Iteration 5: log pseudolikelihood = -3378.7162

    Multinomial logistic regression Number of obs = 2,727
    Wald chi2(69) = 441.80
    Prob > chi2 = 0.0000
    Log pseudolikelihood = -3378.7162 Pseudo R2 = 0.1030

    (Std. Err. adjusted for 2,036 clusters in id)
    ----------------------------------------------------------------------------------------------------------------
    | Robust
    jointcat | RRR Std. Err. z P>|z| [95% Conf. Interval]
    -----------------------------------------------+----------------------------------------------------------------
    4_6 |
    age | 1.036761 .0096243 3.89 0.000 1.018068 1.055797
    bmi | 1.074883 .019771 3.93 0.000 1.036822 1.11434
    dep | 4.243047 1.219357 5.03 0.000 2.415803 7.452365
    |
    edn |
    Trade/apprenticeship | .760332 .2664762 -0.78 0.434 .3825425 1.511217
    Certificate/diploma | 1.136693 .302476 0.48 0.630 .6747411 1.914915
    Bachelor degree or higher | .3718793 .1046642 -3.51 0.000 .2142073 .6456095
    |
    sf1_ | 6.797719 2.269404 5.74 0.000 3.533388 13.07781
    asthma | 2.378593 .6927281 2.98 0.003 1.344067 4.209391
    |
    pa |
    Active but not sufficiently active for health | 1.015569 .2697355 0.06 0.954 .6034343 1.709184
    Sufficiently active for health | 1.593989 .4279161 1.74 0.082 .9418354 2.697712
    |
    sex | 1.32749 .2853553 1.32 0.188 .8710805 2.023039
    diabetes | .849253 .2344682 -0.59 0.554 .4943459 1.45896
    smoker | 1.363192 .2704338 1.56 0.118 .9240458 2.01104
    anxiety | 1.484848 .6610965 0.89 0.375 .6204446 3.553539
    |
    osteo |
    Osteoporosis | 2.712504 .9650428 2.80 0.005 1.350622 5.447624
    Don't know/ not stated | 2.027536 1.48646 0.96 0.335 .4818609 8.531303
    |
    income |
    $12,001 to $20,000 | .6591869 .4399156 -0.62 0.532 .1782173 2.438188
    $20,001 to $40,000 | .7673286 .4694919 -0.43 0.665 .2312979 2.545606
    $40,001 to $60,000 | .8153318 .5277369 -0.32 0.752 .2292874 2.89927
    $60,001 to $80,000 | .7181715 .4778251 -0.50 0.619 .1949367 2.645836
    More than $80,000 | .4474894 .296304 -1.21 0.225 .1222255 1.638339
    Not stated | 1.393372 .9507262 0.49 0.627 .3658261 5.307125
    |
    pension | .7953064 .2309513 -0.79 0.430 .4501442 1.405133
    _cons | .0047684 .0051108 -4.99 0.000 .0005835 .0389674
    -----------------------------------------------+----------------------------------------------------------------
    2_3 |
    age | 1.020224 .0083316 2.45 0.014 1.004024 1.036685
    bmi | 1.051763 .0164636 3.22 0.001 1.019984 1.084531
    dep | 1.653601 .4134177 2.01 0.044 1.013028 2.699232
    |
    edn |
    Trade/apprenticeship | .5892348 .173073 -1.80 0.072 .3313365 1.04787
    Certificate/diploma | 1.176401 .2445243 0.78 0.434 .7827558 1.768008
    Bachelor degree or higher | .4204412 .0980657 -3.71 0.000 .2661748 .6641154
    |
    sf1_ | 4.352846 1.49939 4.27 0.000 2.215985 8.55027
    asthma | 1.407778 .338524 1.42 0.155 .8787158 2.255381
    |
    pa |
    Active but not sufficiently active for health | 1.094137 .2489621 0.40 0.693 .7004665 1.709054
    Sufficiently active for health | 1.393773 .2947881 1.57 0.116 .9207884 2.109718
    |
    sex | 1.107256 .1914985 0.59 0.556 .7889219 1.55404
    diabetes | .954024 .2283128 -0.20 0.844 .5968339 1.524983
    smoker | 1.113897 .1878848 0.64 0.523 .8003321 1.550316
    anxiety | 1.227461 .4695717 0.54 0.592 .579931 2.598001
    |
    osteo |
    Osteoporosis | 1.236368 .3336241 0.79 0.432 .7285473 2.098156
    Don't know/ not stated | 3.147137 1.892815 1.91 0.057 .9682086 10.22969
    |
    income |
    $12,001 to $20,000 | .7871697 .4505604 -0.42 0.676 .2563642 2.417015
    $20,001 to $40,000 | .877213 .4572188 -0.25 0.802 .3158231 2.436499
    $40,001 to $60,000 | 1.193452 .6441708 0.33 0.743 .4143481 3.437515
    $60,001 to $80,000 | 1.024806 .5804651 0.04 0.965 .3376824 3.110102
    More than $80,000 | 1.268613 .6803389 0.44 0.657 .443447 3.629245
    Not stated | 1.636346 .970103 0.83 0.406 .5119677 5.23007
    |
    pension | .9057467 .2157995 -0.42 0.678 .56781 1.444809
    _cons | .0702752 .0672265 -2.78 0.006 .0107777 .4582232
    -----------------------------------------------+----------------------------------------------------------------
    1 |
    age | 1.014911 .008406 1.79 0.074 .998569 1.031521
    bmi | 1.015144 .0174174 0.88 0.381 .9815745 1.049862
    dep | 1.989391 .5843013 2.34 0.019 1.118703 3.537736
    |
    edn |
    Trade/apprenticeship | .8306563 .2788478 -0.55 0.580 .4302062 1.603859
    Certificate/diploma | .826352 .1859039 -0.85 0.397 .5317052 1.284279
    Bachelor degree or higher | .5925781 .1424129 -2.18 0.029 .3699801 .9491018
    |
    sf1_ | 1.818395 .5769021 1.88 0.059 .9764203 3.38641
    asthma | 1.487441 .3748721 1.58 0.115 .9076445 2.437607
    |
    pa |
    Active but not sufficiently active for health | 1.304954 .3629822 0.96 0.339 .7565325 2.250935
    Sufficiently active for health | 1.030397 .2738413 0.11 0.910 .6120501 1.734692
    |
    sex | .9755547 .1829759 -0.13 0.895 .6754598 1.408976
    diabetes | .8844831 .2152513 -0.50 0.614 .5489563 1.425087
    smoker | .9595969 .1715418 -0.23 0.818 .675965 1.362239
    anxiety | 1.290782 .5440507 0.61 0.545 .5650422 2.948664
    |
    osteo |
    Osteoporosis | .8513535 .2707415 -0.51 0.613 .4564747 1.587827
    Don't know/ not stated | 2.197583 1.404351 1.23 0.218 .6280444 7.689537
    |
    income |
    $12,001 to $20,000 | 1.620713 1.021803 0.77 0.444 .4710355 5.576462
    $20,001 to $40,000 | 1.342267 .7535038 0.52 0.600 .4466866 4.033434
    $40,001 to $60,000 | 1.539405 .8804714 0.75 0.451 .5017686 4.722831
    $60,001 to $80,000 | 1.244647 .7473012 0.36 0.715 .3836848 4.03755
    More than $80,000 | 1.466208 .8338315 0.67 0.501 .4809701 4.469648
    Not stated | 1.030301 .6460677 0.05 0.962 .301442 3.521474
    |
    pension | .7899527 .2029652 -0.92 0.359 .4774182 1.307083
    _cons | .2227356 .2165704 -1.54 0.122 .0331245 1.49772
    -----------------------------------------------+----------------------------------------------------------------
    0 | (base outcome)
    ----------------------------------------------------------------------------------------------------------------
    Note: _cons estimates baseline relative risk for each outcome.


  • #2
    The generalized ordered logit model obtained with the npl option in -gologit2- has the same number of parameters as the multinomial logit model of -mlogit- but these parameters estimate different things. I'd have to refresh my memory with some reading in order to give a deeper answer, but (roughly): For a response with j = 1, ... k categories, in the multinomial model the coefficients will reflect the probability of the response being equal to j vs. equal to a fixed base category, while in the generalized ordered logit model, they will reflect the probability of the response being greater than j vs. less than or equal to j. One place I'd start to get into this more deeply would be the Stata Journal article documenting gologit2.

    However, I believe that -mlogit- and -gologit2- will yield similar (though not in general identical) predicted probabilities for individuals.

    (By the way: If you enclose your output here with "Code Tags" (the "#" icon), the spacing is preserved and it will be easier to read.)
    Last edited by Mike Lacy; 27 Oct 2022, 07:34. Reason: Inserted "less than *or equal*"

    Comment


    • #3
      Many thanks for your reply Mike. I accept that the models are estimating different things, but that still doesn't explain why they produce opposite conclusions. ORs for depression (dep) are significantly less than 1 in the gologit2 model, and RRRs are significantly greater than 1 in the mlogit model. I would have expected the substantive results to be similar.

      Comment


      • #4
        Whoops, very sorry. Please ignore my last post. It is an interpretation problem, I have figured it out now.

        Comment

        Working...
        X