Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: Midas


From   Roger Harbord <[email protected]>
To   [email protected], [email protected]
Subject   Re: st: Midas
Date   Thu, 5 May 2011 21:24:39 +0100

- metandi tp fn fn tn- gives the same error message  (Stata 11.2;
-metandi- user-written command from SJ-9-2  st0163).
- metandi tp fn fn tn, gllamm- runs without error, however. In my
experience, -gllamm- is somewhat more reliable (if slower) than
-xtmelogit-, which is used by -midas- and -metandi- by default when
available (i.e. in Stata 10 and above).

In this case, some of the standard errors and confidence intervals are
missing, and the correlation between the logit-transformed sensitivity
and specificity is estimated as -1, on the boundary of the parameter
space. This may be related to the convergence issues; see Riley et al
2007 <http://dx.doi.org/10.1186/1471-2288-7-3>

. metandi tp fp fn tn, gllamm plot

True  positives: tp                   False positives: fp
False negatives: fn                   True  negatives: tn

Fitting univariate model for Sensitivity, Se = tp / (tp + fn)

Fitting univariate model for Specificity, Sp = tn / (tn + fp)

Fitting bivariate model:

Running adaptive quadrature
Iteration 0:    log likelihood = -57.666576
Iteration 1:    log likelihood = -56.600815
Iteration 2:    log likelihood = -56.386241
Iteration 3:    log likelihood =  -56.38023
Iteration 4:    log likelihood = -56.380145
Iteration 5:    log likelihood = -56.380145


Adaptive quadrature has converged, running Newton-Raphson
Iteration 0:   log likelihood = -56.380145
Iteration 1:   log likelihood = -56.380145  (backed up)
Iteration 2:   log likelihood = -56.380129
Iteration 3:   log likelihood = -56.380129

Meta-analysis of diagnostic accuracy

Log likelihood   = -56.380129                     Number of studies =       11
------------------------------------------------------------------------------
            |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
Bivariate    |
 E(logitSe) |   2.219334    .362558                      1.508733    2.929934
 E(logitSp) |   .5358204    .214943                      .1145398     .957101
Var(logitSe) |   .8619995   .5538091                      .2447008    3.036538
Var(logitSp) |   .3491799    .224272                      .0991603    1.229591
Corr(logits) |         -1          .                             .           .
-------------+----------------------------------------------------------------
HSROC        |
     Lambda |   2.442185   .2538694                       1.94461     2.93976
      Theta |   .5494571   .2833487                      -.005896     1.10481
       beta |  -.4518337   .3683124    -1.23   0.220    -1.173713    .2700453
    s2alpha |          0          .                             .           .
    s2theta |   .5486282          .                             .           .
-------------+----------------------------------------------------------------
Summary pt.  |
         Se |   .9019723   .0320568                      .8188733    .9493065
         Sp |   .6308396   .0500561                      .5286037     .722541
        DOR |   15.72346   4.089394                       9.44421    26.17765
        LR+ |   2.443307    .276928                      1.956598    3.051086
        LR- |   .1553925   .0430096                      .0903304    .2673167
      1/LR- |   6.435319   1.781172                      3.740882    11.07047
------------------------------------------------------------------------------
Covariance between estimates of E(logitSe) & E(logitSp)   -.055003

Roger.
--
Roger Harbord
http://www.epi.bris.ac.uk/staff/rharbord.htm



On 5 May 2011 13:07, Ronan Conroy <[email protected]> wrote:
> On 2011 Beal 5, at 09:17, <[email protected]> <[email protected]> wrote:
>
>> When I run the midas command
>>
>> midas tp fp fn tn
>>
>> with this data:
>>
>> tp    fp      fn      tn
>> 17    27      1       47
>> 18    3       2       11
>> 21    5       1       13
>> 44    3       22      15
>> 17    13      0       12
>> 56    28      1       16
>> 23    16      0       14
>> 79    9       23      26
>> 24    1       6       17
>> 28    21      3       24
>> 66    38      12      63
>
>
> The problem is reproducible on StataSE 11.2 on OS X.
>
> The sting is in the line
>
> 23      16      0       14
>
> If you omit this, midas runs.
>
> And that's all the help I can give, I'm afraid.
>

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index