Announcement

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

  • Cure models

    I'm trying to fit a split population survival model using CUREGR to estimate the proportion of drug traffickers who commit no further offence. When I go to check the residuals, however, I get the following message: 'varlist not allowed.' The code I'm trying to run and output I get appears below:

    stset reoff1xfreetime, failure(reoff1xflag==1) scale(1)
    cureregr i.agegroup sex i.disadvantage if contactrank==1, dist(lognormal) class(mixture) link(linear) scale(i.agegroup sex i.disadvantage)
    estat ic
    predict double cs, csnell
    stset cs, failure(reoff1xflag)
    sts generate km=s
    generate double H=-ln(km)
    line H cs cs, sort


    . do "C:\Users\dwath\AppData\Local\Temp\STD1520_000000. tmp"

    . stset reoff1xfreetime, failure(reoff1xflag==1) scale(1)

    failure event: reoff1xflag == 1
    obs. time interval: (0, reoff1xfreetime]
    exit on or before: failure


    487,695 total observations
    23,998 observations end on or before enter()

    463,697 observations remaining, representing
    298,284 failures in single-record/single-failure data
    455749164 total analysis time at risk and under observation
    at risk from t = 0
    earliest observed entry t = 0
    last observed exit t = 10,831

    . cureregr i.agegroup sex i.disadvantage if contactrank==1, dist(lognormal) class(mixture) link(linear) scale(i.agegroup sex i.disadvantage)

    failure _d: reoff1xflag == 1
    analysis time _t: reoff1xfreetime
    cf: linear, kn: lognormal, model: mixture
    cf_initial_coef: 0.3506 pi: 0.3506

    Fitting constant-only model:

    Iteration 0: log likelihood = -828933.25 (not concave)
    Iteration 1: log likelihood = -277200.84
    Iteration 2: log likelihood = -269109.38
    Iteration 3: log likelihood = -266528.91
    Iteration 4: log likelihood = -266363.71
    Iteration 5: log likelihood = -266360.44
    Iteration 6: log likelihood = -266360.44

    Fitting full model:

    Iteration 0: log likelihood = -266360.44 (not concave)
    Iteration 1: log likelihood = -265145.47
    Iteration 2: log likelihood = -264766.41
    Iteration 3: log likelihood = -264620.63
    Iteration 4: log likelihood = -264616.74
    Iteration 5: log likelihood = -264616.73

    No. of subjects = 60919 Number of obs = 60,919
    LR chi2(18) = 3487.41
    Log likelihood = -264616.73 Prob > chi2 = 0.0000


    Coef. Std. Err. z P>z [95% Conf. Interval]

    cure_frac
    agegroup
    20- .0978393 .01061 9.22 0.000 .0770442 .1186345
    25- .1417434 .0111299 12.74 0.000 .1199291 .1635577
    30- .1664177 .0094989 17.52 0.000 .1478002 .1850353
    40- .2424493 .0114263 21.22 0.000 .2200541 .2648444
    50- .4224677 .0167158 25.27 0.000 .3897054 .45523

    sex -.1218376 .0090675 -13.44 0.000 -.1396096 -.1040656

    disadvantage
    advantaged -.0394597 .0124852 -3.16 0.002 -.0639302 -.0149892
    disadvantaged -.0385665 .012088 -3.19 0.001 -.0622585 -.0148745
    highly disadvantaged -.0793836 .0118008 -6.73 0.000 -.1025128 -.0562545

    _cons .3399919 .0211807 16.05 0.000 .2984785 .3815054

    scale
    agegroup
    20- -.4246738 .0476752 -8.91 0.000 -.5181154 -.3312321
    25- -.4220895 .0502865 -8.39 0.000 -.5206493 -.3235298
    30- -.4657804 .0445832 -10.45 0.000 -.5531618 -.3783989
    40- -.6488177 .0526931 -12.31 0.000 -.7520942 -.5455412
    50- -1.093875 .0808464 -13.53 0.000 -1.252331 -.9354194

    sex -.0100778 .0424117 -0.24 0.812 -.0932033 .0730477

    disadvantage
    advantaged .2256552 .0543107 4.15 0.000 .1192083 .3321022
    disadvantaged .1756349 .0527449 3.33 0.001 .0722568 .2790129
    highly disadvantaged .2866025 .0513186 5.58 0.000 .1860198 .3871851

    _cons -6.705231 .0989532 -67.76 0.000 -6.899176 -6.511286

    shape
    _cons -.7465485 .0069688 -107.13 0.000 -.7602071 -.73289


    . estat ic

    Akaike's information criterion and Bayesian information criterion


    Model N ll(null) ll(model) df AIC BIC

    . 60,919 -266360.4 -264616.7 21 529275.5 529464.8

    Note: BIC uses N = number of observations. See [R] BIC note.

    . predict double cs, csnell
    varlist not allowed
    r(101);

    end of do-file

    r(101);

    .



  • #2
    -ssc desc cureregr- show cureregr is a Stata 13 program providing a predict command cureregr_p used when calling predict after cureregr. Follow the syntax from the cureregr help file examples, or:
    Code:
    version 13 : predict double cs, csnell

    Comment

    Working...
    X