Hi,
I desperately need brains to help me eliminate options here. I have dyads (say, parent-child) that I can follow through at most 7 censuses (1851-1911). Most of the time parents have more than one children, and children can be grouped with each of their parents in several occurrences. Moreover, adults can be both parent and child in different dyads for the same census year. So, individuals can be part of multiple dyads for each census that they are enumerated in. A dyad exists only for a specific year when both individuals are listed in the census. Overall I have more than 2M observations. Sample below.
My goal: binary outcome (coresidence) but taking into account the fact that individuals may be part of multiple dyads for several occurrences. I want to test for effects of dyads (and compare to other dyads like siblings, gr-parents and gr-children, uncles/nephews, etc.). IVs are simple: gender, marital status, SES, rural-urban, among others. I suspect a potential interest to stratify by gender as women married in their parish, but left to the man's one. Women may live farther then men in average.
I read about APIM model, SEM, Social relations model, cross-classified models, I enrolled in the LEMMA courses and bought MLWin to try MCMC models, but still facing a mountain of doubts. Can't figure how to model my data but it seems it is a mix of multiple membership with repeated measures for dyadic data. I could also have gone for the long shot and model all the family members at once (a different model I suppose), but I wanted to include extended family members like cousins as well as half siblings (which can be part of more than one family and "family" is not defined here). I read about margins to compare across models (and ease of interpretation) and to get there I was trying to fit this (simplified):
Results are aggregated, it is difficult to identify which characteristics of either members of the dyads have an effect on the outcome. It may be enough though to compare between types of dyads. But Mr. Schecter suggested the addition of interactions in a previous thread, maybe there is more to extract from the data.
Any suggestion or model elimination is welcome.
I desperately need brains to help me eliminate options here. I have dyads (say, parent-child) that I can follow through at most 7 censuses (1851-1911). Most of the time parents have more than one children, and children can be grouped with each of their parents in several occurrences. Moreover, adults can be both parent and child in different dyads for the same census year. So, individuals can be part of multiple dyads for each census that they are enumerated in. A dyad exists only for a specific year when both individuals are listed in the census. Overall I have more than 2M observations. Sample below.
My goal: binary outcome (coresidence) but taking into account the fact that individuals may be part of multiple dyads for several occurrences. I want to test for effects of dyads (and compare to other dyads like siblings, gr-parents and gr-children, uncles/nephews, etc.). IVs are simple: gender, marital status, SES, rural-urban, among others. I suspect a potential interest to stratify by gender as women married in their parish, but left to the man's one. Women may live farther then men in average.
I read about APIM model, SEM, Social relations model, cross-classified models, I enrolled in the LEMMA courses and bought MLWin to try MCMC models, but still facing a mountain of doubts. Can't figure how to model my data but it seems it is a mix of multiple membership with repeated measures for dyadic data. I could also have gone for the long shot and model all the family members at once (a different model I suppose), but I wanted to include extended family members like cousins as well as half siblings (which can be part of more than one family and "family" is not defined here). I read about margins to compare across models (and ease of interpretation) and to get there I was trying to fit this (simplified):
Code:
melogit coresid i.ego_sex i.census i.agegrp || dyad:, covariance(unstructured) vce(cluster dyad)
Any suggestion or model elimination is welcome.
Code:
* Example generated by -dataex-. To install: ssc install dataex clear input long ego int census byte(ligne2 head kin2 half coresid same_sdistnam same_distnam same_context ego_sex ego_age agediff ego_ses ego_mat ego_context) long(ego_serial dyad) byte kin_lsj float(prop agegrp oldest) byte _est_m1 1000019 6 3 1 1 0 0 1 0 0 2 18 29 0 1 5 237801 248 0 0 0 0 1 936904 6 3 1 2 0 0 1 0 0 1 47 29 0 1 5 237802 248 0 .29 20 1 1 1000019 6 3 1 1 0 0 1 0 0 2 18 23 0 1 5 237801 265 0 0 0 0 1 976687 6 3 1 2 0 0 1 0 0 2 41 23 0 1 5 237802 265 0 .29 20 1 1 1000026 6 3 0 1 0 1 0 0 0 1 19 28 0 6 5 237802 539 0 .29 0 0 1 936904 6 3 1 2 0 1 0 0 0 1 47 28 0 1 5 237802 539 0 .29 20 1 1 1000026 6 3 0 1 0 1 0 0 0 1 19 22 0 6 5 237802 556 0 .29 0 0 1 976687 6 3 1 2 0 1 0 0 0 2 41 22 0 1 5 237802 556 0 .29 20 1 1 28866 6 3 0 1 0 1 0 0 0 2 0 22 0 6 5 237989 675 0 0 0 0 1 1000027 6 3 1 2 0 1 0 0 0 2 22 22 0 1 5 237989 675 0 0 20 1 1 1000027 6 3 1 1 0 0 1 0 0 2 22 25 0 1 5 237989 818 0 0 20 0 1 936904 6 3 1 2 0 0 1 0 0 1 47 25 0 1 5 237802 818 0 .29 20 1 1 1000027 6 3 1 1 0 0 1 0 0 2 22 19 0 1 5 237989 835 0 0 20 0 1 976687 6 3 1 2 0 0 1 0 0 2 41 19 0 1 5 237802 835 0 .29 20 1 1 1000094 6 3 0 1 0 1 0 0 0 1 12 35 0 6 5 237802 2076 0 .29 0 0 1 936904 6 3 1 2 0 1 0 0 0 1 47 35 0 1 5 237802 2076 0 .29 20 1 1 1000094 6 3 0 1 0 1 0 0 0 1 12 29 0 6 5 237802 2093 0 .29 0 0 1 976687 6 3 1 2 0 1 0 0 0 2 41 29 0 1 5 237802 2093 0 .29 20 1 1 1000095 6 3 0 1 0 1 0 0 0 2 4 43 0 6 5 237802 2343 0 .29 0 0 1 936904 6 3 1 2 0 1 0 0 0 1 47 43 0 1 5 237802 2343 0 .29 20 1 1 1000095 6 3 0 1 0 1 0 0 0 2 4 37 0 6 5 237802 2360 0 .29 0 0 1 976687 6 3 1 2 0 1 0 0 0 2 41 37 0 1 5 237802 2360 0 .29 20 1 1 1000102 6 3 0 1 0 1 0 0 0 2 14 33 0 6 5 237802 2640 0 .29 0 0 1 936904 6 3 1 2 0 1 0 0 0 1 47 33 0 1 5 237802 2640 0 .29 20 1 1 1000102 6 3 0 1 0 1 0 0 0 2 14 27 0 6 5 237802 2657 0 .29 0 0 1 976687 6 3 1 2 0 1 0 0 0 2 41 27 0 1 5 237802 2657 0 .29 20 1 1 1000110 6 3 0 1 0 1 0 0 0 2 16 31 0 6 5 237802 2925 0 .29 0 0 1 936904 6 3 1 2 0 1 0 0 0 1 47 31 0 1 5 237802 2925 0 .29 20 1 1 1000110 6 3 0 1 0 1 0 0 0 2 16 25 0 6 5 237802 2942 0 .29 0 0 1 976687 6 3 1 2 0 1 0 0 0 2 41 25 0 1 5 237802 2942 0 .29 20 1 1 1001254 1 3 0 1 0 1 0 0 0 2 0 23 0 6 2 242426 20815 0 0 0 0 1 1001308 1 3 1 2 0 1 0 0 0 1 23 23 3 1 2 242426 20815 0 0 20 1 1 1001254 1 3 0 1 0 1 0 0 0 2 0 19 0 6 2 242426 21116 0 0 0 0 1 951044 1 3 1 2 0 1 0 0 0 2 19 19 0 1 2 242426 21116 0 0 0 1 1 1003046 6 3 0 1 0 1 0 0 0 2 3 35 0 6 2 241142 22957 0 0 0 0 1 1001322 6 3 1 2 0 1 0 0 0 1 38 35 2 1 2 241142 22957 0 0 20 1 1 1003047 6 3 0 1 0 1 0 0 0 2 0 38 0 6 2 241142 22958 0 0 0 0 1 1001322 6 3 1 2 0 1 0 0 0 1 38 38 2 1 2 241142 22958 0 0 20 1 1 1003054 6 3 0 1 0 1 0 0 0 2 4 34 0 6 2 241142 22959 0 0 0 0 1 1001322 6 3 1 2 0 1 0 0 0 1 38 34 2 1 2 241142 22959 0 0 20 1 1 1001438 5 3 0 1 0 1 0 0 0 2 5 33 0 6 5 236055 26578 0 0 0 0 1 1001513 5 3 1 2 0 1 0 0 0 1 38 33 1 1 5 236055 26578 0 0 20 1 1 1001438 5 3 0 1 0 1 0 0 0 2 5 25 0 6 5 236055 26621 0 0 0 0 1 934319 5 3 1 2 0 1 0 0 0 2 30 25 0 1 5 236055 26621 0 0 20 1 1 617778 5 3 0 1 0 1 0 0 0 1 5 28 0 6 4 239033 27259 0 .25 0 0 1 1001460 5 3 1 2 0 1 0 0 0 2 33 28 0 1 4 239033 27259 0 .25 20 1 1 617785 5 3 0 1 0 1 0 0 0 1 9 24 0 6 4 239033 27261 0 .25 0 0 1 1001460 5 3 1 2 0 1 0 0 0 2 33 24 0 1 4 239033 27261 0 .25 20 1 1 617786 5 3 0 1 0 1 0 0 0 1 4 29 0 6 4 239033 27262 0 .25 0 0 1 1001460 5 3 1 2 0 1 0 0 0 2 33 29 0 1 4 239033 27262 0 .25 20 1 1 617793 4 3 0 1 0 1 0 0 0 1 5 18 0 6 4 232847 27264 0 0 0 0 1 1001460 4 3 1 2 0 1 0 0 0 2 23 18 0 1 4 232847 27264 0 0 20 1 1 617793 5 3 0 1 0 1 0 0 0 1 15 18 0 6 4 239033 27264 0 .25 0 0 1 1001460 5 3 1 2 0 1 0 0 0 2 33 18 0 1 4 239033 27264 0 .25 20 1 1 617847 4 3 0 1 0 1 0 0 0 1 3 20 0 6 4 232847 27266 0 0 0 0 1 1001460 4 3 1 2 0 1 0 0 0 2 23 20 0 1 4 232847 27266 0 0 20 1 1 617847 5 3 0 1 0 1 0 0 0 1 13 20 0 6 4 239033 27266 0 .25 0 0 1 1001460 5 3 1 2 0 1 0 0 0 2 33 20 0 1 4 239033 27266 0 .25 20 1 1 617862 5 3 0 1 0 1 0 0 0 1 2 31 0 6 4 239033 27269 0 .25 0 0 1 1001460 5 3 1 2 0 1 0 0 0 2 33 31 0 1 4 239033 27269 0 .25 20 1 1 1001498 2 3 0 1 0 1 0 0 0 2 6 32 0 6 4 200062 27996 0 0 0 0 1 5157720 2 3 1 2 0 1 0 0 0 1 38 32 3 1 4 200062 27996 0 0 20 1 1 1001498 3 3 0 1 0 1 0 0 0 2 16 32 0 6 4 230195 27996 0 .38 0 0 1 5157720 3 3 1 2 0 1 0 0 0 1 48 32 3 1 4 230195 27996 0 .38 20 1 1 1001498 4 3 0 1 0 1 0 0 0 2 26 32 0 6 4 231640 27996 0 .25 20 0 1 5157720 4 3 1 2 0 1 0 0 0 1 58 32 3 1 4 231640 27996 0 .25 50 1 1 1001498 5 3 0 1 0 1 0 0 0 2 36 33 0 6 4 234254 27996 0 0 20 0 1 5157720 5 3 0 2 0 1 0 0 0 1 69 33 0 1 4 234254 27996 0 0 50 1 1 1001498 2 3 0 1 0 1 0 0 0 2 6 34 0 6 4 200062 27997 0 0 0 0 1 5157721 2 3 1 2 0 1 0 0 0 2 40 34 0 1 4 200062 27997 0 0 20 1 1 1001498 3 3 0 1 0 1 0 0 0 2 16 33 0 6 4 230195 27997 0 .38 0 0 1 5157721 3 3 1 2 0 1 0 0 0 2 49 33 0 1 4 230195 27997 0 .38 20 1 1 1001498 4 3 0 1 0 1 0 0 0 2 26 33 0 6 4 231640 27997 0 .25 20 0 1 5157721 4 3 1 2 0 1 0 0 0 2 59 33 0 1 4 231640 27997 0 .25 50 1 1 1001498 5 3 0 1 0 1 0 0 0 2 36 34 0 6 4 234254 27997 0 0 20 0 1 5157721 5 3 0 2 0 1 0 0 0 2 70 34 0 1 4 234254 27997 0 0 50 1 1 1001499 3 3 0 1 0 1 0 0 0 1 7 41 0 6 4 230195 28147 0 .38 0 0 1 5157720 3 3 1 2 0 1 0 0 0 1 48 41 3 1 4 230195 28147 0 .38 20 1 1 1001499 4 3 0 1 0 1 0 0 0 1 16 42 3 6 4 231640 28147 0 .25 0 0 1 5157720 4 3 1 2 0 1 0 0 0 1 58 42 3 1 4 231640 28147 0 .25 50 1 1 1001499 5 3 0 1 0 1 0 0 0 1 27 42 0 6 4 234254 28147 0 0 20 0 1 5157720 5 3 0 2 0 1 0 0 0 1 69 42 0 1 4 234254 28147 0 0 50 1 1 1001499 3 3 0 1 0 1 0 0 0 1 7 42 0 6 4 230195 28148 0 .38 0 0 1 5157721 3 3 1 2 0 1 0 0 0 2 49 42 0 1 4 230195 28148 0 .38 20 1 1 1001499 4 3 0 1 0 1 0 0 0 1 16 43 3 6 4 231640 28148 0 .25 0 0 1 5157721 4 3 1 2 0 1 0 0 0 2 59 43 0 1 4 231640 28148 0 .25 50 1 1 1001499 5 3 0 1 0 1 0 0 0 1 27 43 0 6 4 234254 28148 0 0 20 0 1 5157721 5 3 0 2 0 1 0 0 0 2 70 43 0 1 4 234254 28148 0 0 50 1 1 1001520 5 3 0 1 0 1 0 0 0 1 6 32 0 6 5 236055 28228 0 0 0 0 1 1001513 5 3 1 2 0 1 0 0 0 1 38 32 1 1 5 236055 28228 0 0 20 1 1 1001520 5 3 0 1 0 1 0 0 0 1 6 24 0 6 5 236055 28349 0 0 0 0 1 934319 5 3 1 2 0 1 0 0 0 2 30 24 0 1 5 236055 28349 0 0 20 1 1 1001567 2 3 0 1 0 1 0 0 0 2 9 29 0 6 4 200062 28913 0 0 0 0 1 5157720 2 3 1 2 0 1 0 0 0 1 38 29 3 1 4 200062 28913 0 0 20 1 1 1001567 3 3 0 1 0 1 0 0 0 2 19 29 0 6 4 230195 28913 0 .38 0 0 1 5157720 3 3 1 2 0 1 0 0 0 1 48 29 3 1 4 230195 28913 0 .38 20 1 1 1001567 4 3 0 1 0 1 0 0 0 2 29 29 0 6 4 231640 28913 0 .25 20 0 1 5157720 4 3 1 2 0 1 0 0 0 1 58 29 3 1 4 231640 28913 0 .25 50 1 1 1001567 5 3 1 1 0 0 0 1 0 2 39 30 0 1 5 235033 28913 0 0 20 0 1 5157720 5 3 0 2 0 0 0 1 0 1 69 30 0 1 4 234254 28913 0 0 50 1 1 end label values census census label def census 1 "1861", modify label def census 2 "1871", modify label def census 3 "1881", modify label def census 4 "1891", modify label def census 5 "1901", modify label def census 6 "1911", modify label values ligne2 lineage label def lineage 3 "Neutral", modify label values head head label def head 0 "No", modify label def head 1 "Yes", modify label values kin2 role label def role 1 "Parent", modify label def role 2 "Children", modify label values half half label def half 0 "No", modify label values coresid coresid label def coresid 0 "No", modify label def coresid 1 "Yes", modify label values same_sdistnam sdistnam label def sdistnam 0 "No", modify label def sdistnam 1 "Yes", modify label values same_distnam distnam label def distnam 0 "No", modify label def distnam 1 "Yes", modify label values same_context same_context label def same_context 0 "No", modify label values ego_sex sex label def sex 1 "Male", modify label def sex 2 "Female", modify label values ego_ses ses label def ses 1 "Higher occupations", modify label def ses 2 "Skilled workers", modify label def ses 3 "Farmers", modify label values ego_mat marital label def marital 1 "Married", modify label def marital 6 "Single", modify label values ego_context context label def context 2 "Rural", modify label def context 4 "Saturated rural", modify label def context 5 "Urban", modify label values kin_lsj kinlsj label def kinlsj 0 "No", modify label values agegrp agegrp label def agegrp 0 "0-19", modify label def agegrp 20 "20-49", modify label def agegrp 50 "50+", modify
