Nested case control study

[PRITI SINGH]

Hello ,

I am trying to create a 1:3 matching for a nested case control study (cases could serve as control until 0 for the same length of stay and admission year). What code can I use to match based on adm_year and LOS_cumulative_days without replacement. I tried previous codes using join by but doesnt seem to help. Thank you.

input float(Person_visit_ID adm_year LOS_cumulative_days case)
3 2019 2.5 0
3 2019 9.3 0
7 2021 27.2 0
7 2021 38.4 0
7 2021 .8 0
7 2021 31.9 0
8 2019 6.4 0
8 2019 3 0
10 2020 4.7 0
17 2020 10.9 0
17 2020 7 0
17 2020 7.1 0
21 2019 9.5 0
21 2019 32 0
24 2021 .4 0
24 2021 18.7 0
30 2019 13.2 0
30 2019 4.4 0
31 2019 4.3 0
44 2019 1.7 0
44 2019 .6 0
44 2019 4.8 0
57 2021 4.7 0
62 2019 55.3 0
62 2019 39.3 0
62 2019 56.6 0
62 2019 1.4 0
62 2019 36.5 0
62 2019 54 0
62 2019 15.5 0
62 2019 32.8 0
62 2019 30.6 0
64 2019 2.6 0
64 2019 6.5 0
65 2019 6.8 0
65 2019 3.7 0
65 2019 8.8 0
66 2019 2 0
66 2019 7.8 0
66 2019 12.9 0
66 2019 .2 0
66 2019 15.7 0
66 2019 3.7 0
67 2021 6.5 0
67 2021 1 0
68 2019 7.4 0
69 2019 6.6 0
69 2019 7 0
70 2021 .9 0
70 2021 3.9 0
70 2021 6.8 0
73 2019 17.2 0
73 2019 26.4 0
77 2019 5.3 0
77 2019 3.3 0
91 2019 6.1 0
91 2019 11.5 0
93 2019 2 0
93 2019 11.7 0
93 2019 12 0
93 2019 15.1 0
93 2019 9 0
94 2019 6.2 0
94 2019 12.4 0
94 2019 15.5 0
97 2019 4.6 0
97 2019 2.6 0
98 2021 1.2 0
98 2021 5.7 1
100 2019 1.6 0
100 2019 12.7 0
101 2019 4.1 0
103 2019 5.9 0
103 2019 1.8 0
112 2020 1.5 0
112 2020 2.3 0
112 2020 2.6 0
112 2020 4.2 0
113 2020 4 0
113 2020 1.3 0
114 2021 2.5 0
114 2021 5.5 0
115 2019 10.3 0
115 2019 3.5 0
116 2019 5 0
117 2019 18.8 0
117 2019 6.3 0
119 2019 11.1 0
120 2019 4.9 0
121 2019 8.2 0
122 2019 5.5 0
123 2020 7.8 0
128 2019 9.8 0
128 2019 9 0
128 2019 12.9 1
133 2019 4 0
144 2019 4.8 0
144 2019 3.6 1
145 2019 5.8 0
145 2019 2.8 0
end
[/CODE]

[Clyde Schechter]

You asked a closely related question at https://www.statalist.org/forums/for...t-replacements, and I responded to it there. If you wish, you can adapt the code shown there to do this in your data simply by replacing variable names appropriately.

My basic recommendation, however, is that matching without replacement, a moderately bad, if tolerable, idea with 1:1 matching, is an extremely bad idea with (n>1):1 matching.

[PRITI SINGH]

Thank you, found it. I understand the trade off, but want to see the matched pairs, since I should have enough controls to match from.