st: Script searching data in a kinship matrix

[Ginevra Biino <biino@igm.cnr.it>]

Hello everybody,
in a case-control study I have already sampled cases stratifying for sex 
(0,1) and age (<62y, >=62y). I need to sample a group of controls with the 
same characteristics (which I can easily do with the sample command) plus 
one more: the level of relatedness. Therefore controls should be matched to 
cases for sex (2 strata), age(2 strata) and, relatedness (less than a 
certain level). In particular I need that controls are as least as possible 
related to cases, for example each control should have a kinship coefficient 
less than 0.0156 (i.e. 1/64 as for second cousins) with its matched case. In 
the example data set below, there are the sampled cases [20 cases (where 
disease==1): 5 subjects for each strata] and 200 possible controls 
(disease==0) I have already sampled stratifying for age and sex (50 subjects 
per strata).
id disease age sex
109530 0 65 M
109398 0 65 M
109494 0 65 M
110077 0 71 M
109601 0 66 M
109585 0 66 M
114262 0 73 M
109311 0 63 M
109355 0 64 M
110756 0 78 M
111090 0 81 M
110806 0 78 M
110222 0 72 M
110955 0 80 M
110310 0 73 M
109829 0 68 M
109434 0 64 M
110006 0 70 M
109286 0 63 M
109298 0 63 M
109721 0 67 M
110234 0 72 M
110143 0 71 M
133061 0 78 M
110021 0 69 M
110296 0 73 M
109719 0 67 M
110198 0 72 M
110115 0 71 M
110092 0 70 M
109296 0 63 M
109540 0 66 M
109791 0 68 M
109227 0 62 M
109807 0 68 M
109934 0 69 M
125715 0 73 M
109577 0 66 M
110677 0 76 M
111792 0 89 M
110414 0 74 M
109505 0 65 M
111257 0 82 M
109651 0 66 M
109552 0 66 M
109356 0 64 M
110641 0 76 M
109866 0 69 M
110749 0 78 M
110923 0 79 M
109316 0 63 F
105263 0 70 F
110843 0 78 F
109878 0 68 F
110941 0 79 F
111008 0 79 F
109403 0 64 F
110083 0 70 F
109778 0 68 F
109783 0 68 F
109325 0 63 F
109726 0 67 F
109958 0 69 F
110049 0 70 F
110736 0 77 F
114290 0 74 F
110791 0 78 F
111315 0 83 F
109431 0 64 F
114096 0 75 F
109784 0 68 F
110656 0 77 F
114678 0 74 F
32255 0 88 F
109253 0 63 F
133094 0 62 F
111251 0 82 F
109851 0 68 F
109221 0 62 F
109271 0 63 F
110264 0 72 F
109615 0 66 F
110557 0 75 F
110082 0 71 F
110278 0 72 F
110925 0 79 F
110347 0 73 F
109636 0 67 F
110271 0 72 F
109635 0 66 F
109621 0 66 F
110496 0 75 F
109295 0 63 F
110781 0 78 F
109281 0 62 F
110289 0 73 F
111491 0 85 F
109753 0 67 F
109181 0 62 F
110353 0 73 F
104532 0 36 M
105965 0 51 M
105866 0 50 M
105916 0 51 M
106358 0 56 M
103664 0 23 M
105618 0 48 M
109082 0 61 M
104572 0 36 M
105897 0 50 M
105090 0 41 M
108918 0 59 M
104758 0 38 M
103330 0 19 M
104390 0 34 M
109086 0 61 M
105198 0 43 M
104781 0 39 M
109128 0 61 M
105002 0 40 M
108946 0 60 M
133131 0 51 M
106058 0 52 M
115009 0 48 M
104740 0 38 M
132995 0 37 M
103309 0 18 M
103943 0 28 M
105747 0 49 M
103850 0 26 M
104824 0 38 M
104516 0 35 M
106423 0 56 M
105266 0 44 M
105117 0 41 M
104803 0 39 M
105642 0 48 M
108940 0 59 M
104982 0 41 M
105235 0 43 M
104839 0 39 M
104207 0 32 M
105097 0 42 M
104948 0 40 M
104218 0 31 M
104604 0 36 M
105565 0 47 M
104134 0 41 M
105059 0 41 M
104784 0 39 M
105131 0 42 F
115126 0 33 F
105417 0 45 F
103831 0 26 F
133091 0 35 F
106238 0 54 F
104724 0 38 F
105511 0 46 F
105438 0 46 F
103422 0 29 F
105142 0 42 F
103388 0 18 F
104384 0 34 F
105203 0 42 F
105023 0 41 F
105076 0 41 F
105691 0 48 F
104844 0 39 F
104674 0 37 F
104345 0 34 F
104736 0 39 F
103892 0 27 F
103909 0 27 F
1537 0 33 F
104562 0 36 F
108828 0 24 F
103814 0 26 F
105558 0 47 F
103556 0 23 F
109108 0 61 F
105179 0 43 F
104230 0 32 F
133036 0 45 F
104419 0 36 F
105475 0 46 F
103931 0 28 F
113829 0 48 F
133026 0 32 F
104542 0 35 F
104221 0 31 F
104510 0 36 F
105803 0 50 F
106489 0 57 F
105671 0 48 F
137828 0 24 F
104021 0 29 F
106195 0 54 F
133105 0 43 F
105425 0 45 F
104524 0 35 F
106258 1 55 M
105270 1 44 M
106661 1 59 M
104363 1 33 M
108982 1 60 M
106046 1 52 F
103359 1 18 F
105939 1 51 F
104152 1 31 F
105351 1 44 F
110363 1 73 M
114307 1 77 M
110790 1 78 M
109486 1 65 M
109317 1 64 M
114643 1 74 F
114057 1 75 F
109895 1 69 F
110775 1 77 F
110178 1 71 F

What I do not know is how to solve the relatedness problem. I have already 
computed the kinship coefficients matrix of the extended pedigree to whom 
the cases and controls in the example data belong. I do not provide it here 
because is a 190X190 matrix. As an immediate example, such a kinship matrix 
for 5 subjects (ID: 51, 59, 119, 156 and 178) is like:
   51 59 119 156 178
51 0.500 0.000 0.000 0.000 0.000
59 0.000 0.500 0.250 0.000 0.250
119 0.000 0.250 0.500 0.000 0.250
156 0.000 0.000 0.000 0.500 0.000
178 0.000 0.250 0.250 0.000 0.500


In conclusion I need a script that looks down such kinship matrix searching 
for controls satisfying the relatedness condition and that adds this 
information to my data set (.
The information could be reported in many ways (the simplest to obtain it):
For example, the script may add to the original data set as many columns as 
the maximum number of controls satisfying the relatedness condition for 
their matched cases. In particular in correspondence of each case (rows for 
which disease==1) the new columns should return the ID of the matched 
control satisfying the condition: case-control kinship< 0.0156 ; otherwise a 
zero (if condition is not satisfied), and finally a missing value (if the 
subject's ID is not in the kinship matrix).
One other alternative may be that the script adds to the original data set 
as many rows as the maximum number of controls satisfying the relatedness 
condition for their matched cases. Such new rows (for each  matched case) 
may report the control's ID in the ID column, missing values in the disease, 
age and sex columns, and the ID of the matched case in a new column.
Whatever alternative solution is welcome!
Does anybody can help me?
Ginevra


Ginevra Biino, PhD
Institute of Molecular Genetics, CNR
Via Abbiategrasso, 207
27100 Pavia, Italy
Tel +39 382 546363
Fax +39 382 422286
http://www.igm.cnr.it/
Ginevra Biino, PhD
Institute of Molecular Genetics, CNR
Via Abbiategrasso, 207
27100 Pavia, Italy
Tel +39 382 546363
Fax +39 382 422286
http://www.igm.cnr.it/ 

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