meta analysis STATA 16 Negative INTERVAL CONFIDENCE-

[Tommaso Salvitti]

Hi to everybody i trt to conduct a meta analyis (meta regression) with stata 16


The aim is the prevalence of skin disease in diabetic population

Below the dataset

Study	Group_age	Female_percentage	mean_Age	Prev	se(prev)
1	children	19.2	6	0.03	0.008245
2	adult	2	23	0.26	0.055571
3	adult	3.2	32	0.12	0.031311
4	adult	1.2	54	0.10	0.025624
5	children	5.6	13	0.14	0.008974
6	children	27	15	0.02	0.002012
7	adult	4	27	0.11	0.058451
8	adult	0.3	30	0.02	0.019799
9	children	2	15.3	0.08	0.034056
10	children	1.5	16.9	0.05	0.018043
11	adult	1.6	19	0.10	0.034772
12	adult	3.5	20.2	0.10	0.043243
13	adult	0.2	25	0.06	0.020726
14	adult	13	30.4	0.01	0.012902
15	children	2	14	0.35	0.050502
16	children	2.3	13.2	0.03	0.016659
17	children	8.9	12.0	0.27	0.077528
18	children	16.2	11	0.53	0.048038
19	adult	17.3	37.2	0.62	0.095411
20	adult	45.3	35.6	0.02	0.019799
21	children	6	12	0.04	0.009234
22	children	74.3	11	0.09	0.040151
23	adult	15	39	0.05	0.001645
24	adult	16.2	42.3	0.01	0.000831
25	children	3	15.6	0.02	0.002015
26	adult	2.5	18.9	0.16	0.029964
27	adult	6.2	21	0.17	0.030429

My effect size data are present in dataset : prev (prevalence of skin disease in diabetic population) and se(prev) standard error ef the effect size.
so I use this code

meta set prev se(prev)
meta summarize
meta forestplot

meta regress Female_percentage
meta funnelplot
meta bias

My problem is that when i run the instrucion meta summarize I obtain negative confidence interval.
How to solve this problem with STATA 16?


Thnaks to everybody in advance

[Joseph Coveney]

The command wants something symmetric, and so you can transform your prevalence values into log odds, compute the confidence bounds in that metric and enter the prevalence information in that manner into meta set.

If the negative lower confidence bounds bothers you for esthetics alone, then you could just relax your confidence level or set any negative values equal to zero (boundary of the parameter space) when you go to report them.

[Weiwen Ng]

Adding a bit to what Joseph said. Remember that one of the disadvantages of linear probability models, where you apply OLS regression to a binary outcome, is that they can produce predictions or CIs that are out of bounds. By itself, that doesn’t mean that LPMs are invalid and the researcher should be tarred and feathered. It’s just a disadvantage. For that and other reasons, logit or probit regression is preferred. A similar deal occurs with fractional outcomes, where you apply OLS to something that’s a fraction and thus is bounded between 0 and 1. There are actually alternative regression models, e.g. beta, fractional logit, but they’re less well known.

Here, the meta analysis is applying the equivalent of OLS to a fractional outcome. The overall prevalence is low enough to cause the lower CI to be below 0. I think the simplest thing to do is just to truncate the lower bound at 0. I think that people viewing the analysis are either just going to look at the measure of central tendency and not really bother with the issue with CIs, or they would understand anyway. Transforming to log odds is also technically correct. It may produce an asymmetric CI when transformed to the probability scale, and that might confuse some less sophisticated readers.

the latter may sound a bit condescending. However, I really do mean that if you don’t know what a log transformation does, you would probably find an asymmetric CI odd and confusing, and it might distract you from the rest of the analysis. I think many of us know how picky researchers can be, and because we’re all smart we sometimes get hung up on certain things.

[Tommaso Salvitti]

Joseph Coveney and Weiwen Ng thanks a lot...And sorry if i answer so late.

Another question

I'm using stata 16
Is it correct the code below to perform a multivariable meta regression?

meta regress var1 var2 var3

Many thanks in advanced

[Tommaso Salvitti]

Hi to everybody and Hi Weiwen Ng and Joseph Coveney

* Example generated by -dataex-. For more info, type help dataex
clear
input str22 Gruppo byte Study str8 Typeofstudytrialdesign long Totalautism int Female double(Female_percentage Age ID_percentage) long Casis double(prev_ADHD se_prev_ADHD)
"Mixed"                   1 "Pop Reg"    414   79 19.082125603864732 16.28                  .    74    .178743961352657  .018830194533401432
"Children-Adolescent<18"  2 "Clin Com"    94   17 18.085106382978726     9                  .    51   .5425531914893617   .05138395580107374
"Mixed"                   3 "Pop Reg"  20194 4271 21.149846489056156 12.54 10.364464692482915  2897   .1434584530058433  .002466755541035261
"Children-Adolescent<18"  4 "Clin Com"    60   23 38.333333333333336  8.36                 60    19  .31666666666666665   .06005398805642704
"Adult"                   5 "Clin Com"   113    .                  .     .                  .    47    .415929203539823  .046366384411396705
"Children-Adolescent<18"  6 "Clin Com"   883  154 17.440543601359003   8.3              33.64    62  .07021517553793885  .008598572708216004
"Children-Adolescent<18"  7 "Clin Com"   201   33 16.417910447761194  9.13                  .   156   .7761194029850746   .02940183729453583
"Mixed"                   8 "Pop Reg"    586   91 15.529010238907851   9.3                  .   232  .39590443686006827  .020202224801221257
"Adult"                   9 "Pop Reg"   1507  405 26.874585268745854    29              19.18   167   .1108161911081619  .008086130691068913
"Children-Adolescent<18" 10 "Pop Reg"   8325 1469              17.65   9.7                  .  3039  .36504504504504504  .005276580475790189
"Mixed"                  11 "Clin Com"  4123  797 19.330584525830705 18.39                 13   602   .1460101867572156  .005499348673103253
"Children-Adolescent<18" 12 "Clin Com"    94   11 11.702127659574469   8.5                  .    42  .44680851063829785   .05127840836514522
"Children-Adolescent<18" 13 "Clin Com"   284   42 14.788732394366196   8.3                  .   228   .8028169014084507   .02360934437091792
"Children-Adolescent<18" 14 "Clin Com"   172   36 20.930232558139537   4.2                  .    57   .3313953488372093  .035891670958908016
"Children-Adolescent<18" 15 "Clin Com"    68    .               24.6   7.3                  .    37   .5441176470588235  .060397413415836175
"Adult"                  16 "Clin Com"    28   10 35.714285714285715  26.5                  .     5  .17857142857142858   .07237888244444443
"Mixed"                  17 "Clin Com"    35    6 17.142857142857142  15.1                  0    10   .2857142857142857   .07636035483212125
"Mixed"                  18 "Pop Reg"  28468 8734 30.680061823802163  21.5                  . 13793   .4845089222987214 .0029619859035606706
"Adult"                  19 "Clin Com"    50    0                  0  30.2                  0    14                 .28   .06349803146555018
"Mixed"                  20 "Clin Com"    86    9 10.465116279069768  14.9                  0    38   .4418604651162791  .053550649191937534
"Children-Adolescent<18" 21 "Clin Com"    71   13  18.30985915492958  11.8                  .    22  .30985915492957744  .054880979931595526
"Children-Adolescent<18" 22 "Clin Com"    55    9 16.363636363636363  11.9                  .    17   .3090909090909091   .06231207660686132
"Children-Adolescent<18" 23 "Pop Reg"   3319  564              16.99  10.3              19.55  1503  .45284724314552577  .008640254478041821
"Adult"                  24 "Clin Com"    54   28  51.85185185185185    27                  0    16   .2962962962962963   .06213855502280453
"adults"                 25 "pop reg"   4685 1510 32.230522945570975    67                 43   116 .024759871931696906 .0022702578226601487
"Mixed"                  26 "Clin Com"   107   13               12.3 13.63               9.35    43  .40186915887850466  .047396748382179395
"Children-Adolescent<18" 27 "Pop Reg"  71386    .                  .    10                  . 61616   .8631384305045807 .0012863953484885486
"Children-Adolescent<18" 28 "Pop Reg"   6294    .                  .    10                  .   891   .1415633937082936  .004394062479787525
"Adult"                  29 "Pop Reg"   4562    .                  .  41.5                  .   604   .1323980710214818  .005017916645791416
"Adult"                  30 "Pop Reg"  22253    .                  .    34                  . 11610   .5217274075405564  .003348615545439927
"Adult"                  31 "Clin Com"    48    5 10.416666666666668  25.6                  .    18                .375   .06987712429686843
"Children-Adolescent<18" 32 "Clin Com"   143   18 12.587412587412588    10                  .    96   .6713286713286714   .03928081493863158
"Children-Adolescent<18" 33 "Clin Com"   217   29  13.36405529953917   9.7                  .   181   .8341013824884793  .025252299819458597
"Adult"                  34 "Clin Com"    63   22  34.92063492063492    29                  .    26   .4126984126984127   .06202641946766029
"Children-Adolescent<18" 35 "Clin Com"    77   16  20.77922077922078   8.5                  .    63   .8181818181818182   .04395397985592983
"Children-Adolescent<18" 36 "Clin Com"    79    .                  .     .              30.38    18  .22784810126582278   .04719114699416714
"Children-Adolescent<18" 37 "Clin Com"    49   13  26.53061224489796  11.2                  0    24   .4897959183673469     .071413695124874
"Children-Adolescent<18" 38 "Clin Com"    52    9 17.307692307692307   8.7                  .    43   .8269230769230769  .052462679367767266
"Mixed"                  39 "Clin Com"   114   25 21.929824561403507 13.25                  .    87   .7631578947368421  .039818439089930886
"Children-Adolescent<18" 40 "Pop Reg"   2568  491  19.11993769470405     8               18.3   547  .21300623052959503  .008079496382404645
"Children-Adolescent<18" 41 "Clin Com"    89   15 16.853932584269664     9              59.55    16   .1797752808988764   .04070393643506958
"Children-Adolescent<18" 42 "Clin Com"    50    .                  .     4                 14    20                  .4   .06928203230275509
"Children-Adolescent<18" 43 "Clin Com"    99   21  21.21212121212121  9.37                  .    85   .8585858585858586   .03502036666045565
"Children-Adolescent<18" 44 "Clin Com"    50   12                 24  12.7                  .    19                 .38   .06864400920692205
"Children-Adolescent<18" 45 "Clin Com"    80   20                 25  3.77              93.75    55               .6875   .05182226234930312
"Children-Adolescent<18" 46 "Clin Com"    37   15  40.54054054054054    12                  .     7   .1891891891891892   .06438831518199858
"Mixed"                  47 "Clin Com"    84   15 17.857142857142858  19.5  78.57142857142857     1 .011904761904761904  .011833688064142915
"Children-Adolescent<18" 48 "Clin Com"    35    6 17.142857142857142    14                  .    25   .7142857142857143   .07636035483212125
"Children-Adolescent<18" 49 "Clin Com"    34    5 14.705882352941178  10.2                  0    17                  .5   .08574929257125442
"Adult"                  50 "Clin Com"     3    0                  0  19.7                  0     1   .3333333333333333    .2721655269759087
"Mixed"                  51 "Pop Reg"   5651  963 17.041231640417625    12                  .   421  .07450008847991506 .0034930429958927684
"Children-Adolescent<18" 52 "Clin Com"   213    .                  .     9                  .    46    .215962441314554  .028194717420180173
"Children-Adolescent<18" 53 "Clin Com"    37    7  18.91891891891892   9.9                  .     8  .21621621621621623    .0676770479000758
"Mixed"                  54 "Pop Reg"    108   29 26.851851851851855  30.7                  .    14  .12962962962962962   .03232155981976339
"Mixed"                  55 "Clin Com"    42    4  9.523809523809524  13.9                  0    14   .3333333333333333   .07273929674533079
"Children-Adolescent<18" 56 "Clin Com"   123   22  17.88617886178862 10.62                  .    71   .5772357723577236   .04454235542235718
"Children-Adolescent<18" 57 "Clin Com"    50    4                  8 11.95                  0    23                 .46   .07048404074682439
"Children-Adolescent<18" 58 "Clin Com"   201    .                  .    11                  .    70   .3482587064676617  .033603969488125496
"Adult"                  59 "Clin Com"   474  102 21.518987341772153 30.59                  0    46   .0970464135021097  .013596692878530993
"Children-Adolescent<18" 60 "Pop Reg"    101   44  43.56435643564357   6.7              56.44    60    .594059405940594  .048863603672897535
"Children-Adolescent<18" 61 "Pop Reg"    112   14               12.5  11.5                  .    32   .2857142857142857  .042686736047656916
"Children-Adolescent<18" 62 "Pop Reg"   1874  380 20.277481323372466     8                  .   349  .18623265741728923   .00899276522707784
"Children-Adolescent<18" 63 "Clin Com"   197   31 15.736040609137056  9.12                  .    29  .14720812182741116   .02524377499739435
"Mixed"                  64 "Pop Reg"   4790   17                .35  17.5                  .   660  .13778705636743216   .00498016490460324
"Children-Adolescent<18" 65 "Pop Reg"   2352  445 18.920068027210885   8.5                  .   408  .17346938775510204  .007807684141992912
"Children-Adolescent<18" 66 "Clin Com"   108    .              20.37 10.95                  0    75   .6944444444444444   .04432532906279343
"Children-Adolescent<18" 67 "Clin Com"    40    4                 10  11.1                  2     9                .225   .06602556323122129
"Children-Adolescent<18" 68 "Clin Com"    94    9  9.574468085106384  9.02                  0    42  .44680851063829785   .05127840836514522
"Mixed"                  69 "Clin Com"    74    9 12.162162162162163    16                  0    29   .3918918918918919   .05674892985790457
"Adult"                  70 "Pop Reg"   1772  506 28.555304740406324    29                  .   146  .08239277652370203  .006531923987783141
"Children-Adolescent<18" 71 "Clin Com"    61   11   18.0327868852459  11.2              59.01    56   .9180327868852459   .03512240634819573
"Children-Adolescent<18" 72 "Pop Reg"   7773 1349 17.354946610060466     8                  .  1461  .18795831725202625  .004431243357458032
end

Here my dataset

My effect size data are present in dataset : prev_ADHD (PREVALENCE ASD with ADHD) and se_prev_ADHD (standard error ef the effect size).


so I use this code

meta set prev_ADHD se_prev_ADHD
meta summarize

My problem is that when i run the instrucion meta summarize I obtain negative confidence interval.
How to solve this problem with STATA 17?

Joseph Coveney suggested me to transform my prevalence values into log odds, compute the confidence bounds in that metric and enter the prevalence information in that manner into meta set.

HOW TO DO THIS TRANSFORMATION??? AND THEN HOW OBTAIN THE VALUE OF THE PREVALENCE???


THANKS TO EVERYBODY