Hello everyone,
I would like to ask for your feedback on whether my approach to deriving a correlation matrix is appropriate.
I am working with a dataset that includes survey weights, and I want to create a correlation matrix of key variables and covariates.
I have been referring to this document for guidance: https://www.stata.com/manuals13/semexample16.pdf
This is the command I used:
This is the result:
For my understanding, SEM treats dummy variables as continuous, ranging from 0 to 1, and calculates means for these variables in the output (e.g., in the first row of the correlation matrix).
I would like to seek your insights on whether my approach is appropriate. If it is, how should I interpret the correlations between two dummy variables, or between a dummy variable and a continuous variable?
If this approach is not suitable, could you kindly share any alternative methods for deriving a correlation matrix that would be appropriate for survey-weighted data and datasets including both continuous and categorical variables?
Thank you very much for your time and guidance.
I would like to ask for your feedback on whether my approach to deriving a correlation matrix is appropriate.
I am working with a dataset that includes survey weights, and I want to create a correlation matrix of key variables and covariates.
I have been referring to this document for guidance: https://www.stata.com/manuals13/semexample16.pdf
This is the command I used:
Code:
svy,subpop(if lfl7spdied == -1 & dementia_community == 1):sem ( <- black_dummy_2015 others_dummy_2015 burden_con_2015 pp_sumscore_2015 ss_sumscore_2015 overall_caregiver_support_2015 gad2_cont_2017_new age_2015 gender_2015 income_middle_dummy_2015 income_highest_dummy_2015 education_nodegree_dummy_2015 education_degree_dummy_2015), standardized
Code:
----------------------------------------------------------------------------------------------------------------------------------
| Linearized
Standardized | Coefficient std. err. t P>|t| [95% conf. interval]
-----------------------------------------------------------------+----------------------------------------------------------------
mean(black_dummy_2015)| .5414392 .0487681 11.10 0.000 .443579 .6392995
mean(others_dummy_2015)| .6343626 .0856316 7.41 0.000 .4625302 .806195
mean(burden_con_2015)| .7306887 .055289 13.22 0.000 .6197431 .8416343
mean(pp_sumscore_2015)| 1.728036 .09571 18.05 0.000 1.53598 1.920092
mean(ss_sumscore_2015)| .6739034 .053707 12.55 0.000 .5661323 .7816744
mean(overall_caregiver_support_2015)| 1.916074 .1180063 16.24 0.000 1.679277 2.152871
mean(gad2_cont_2017_new)| 2.129797 .0886936 24.01 0.000 1.95182 2.307774
mean(age_2015)| 4.576868 .3258994 14.04 0.000 3.922903 5.230833
mean(gender_2015)| 1.516775 .1235758 12.27 0.000 1.268803 1.764748
mean(income_middle_dummy_2015)| .7021273 .058886 11.92 0.000 .5839639 .8202907
mean(income_highest_dummy_2015)| .5947066 .0601872 9.88 0.000 .4739321 .7154811
mean(education_nodegree_dummy_2015)| .5710484 .041928 13.62 0.000 .4869137 .655183
mean(education_degree_dummy_2015)| .7127419 .0675735 10.55 0.000 .5771457 .8483381
-----------------------------------------------------------------+----------------------------------------------------------------
var(black_dummy_2015)| 1 . . .
var(others_dummy_2015)| 1 . . .
var(burden_con_2015)| 1 . . .
var(pp_sumscore_2015)| 1 . . .
var(ss_sumscore_2015)| 1 . . .
var(overall_caregiver_support_2015)| 1 . . .
var(gad2_cont_2017_new)| 1 . . .
var(age_2015)| 1 . . .
var(gender_2015)| 1 . . .
var(income_middle_dummy_2015)| 1 . . .
var(income_highest_dummy_2015)| 1 . . .
var(education_nodegree_dummy_2015)| 1 . . .
var(education_degree_dummy_2015)| 1 . . .
-----------------------------------------------------------------+----------------------------------------------------------------
cov(black_dummy_2015,others_dummy_2015)| -.3434688 .0442519 -7.76 0.000 -.4322668 -.2546708
cov(black_dummy_2015,burden_con_2015)| -.117412 .0518461 -2.26 0.028 -.2214488 -.0133752
cov(black_dummy_2015,pp_sumscore_2015)| -.0012803 .0516276 -0.02 0.980 -.1048787 .1023182
cov(black_dummy_2015,ss_sumscore_2015)| -.0395679 .0519593 -0.76 0.450 -.1438319 .0646961
cov(black_dummy_2015,overall_caregiver_support_2015)| .0541054 .061301 0.88 0.382 -.068904 .1771148
cov(black_dummy_2015,gad2_cont_2017_new)| -.0531249 .0568641 -0.93 0.354 -.1672311 .0609812
cov(black_dummy_2015,age_2015)| -.0975533 .0563254 -1.73 0.089 -.2105784 .0154718
cov(black_dummy_2015,gender_2015)| -.0685557 .0728298 -0.94 0.351 -.2146995 .077588
cov(black_dummy_2015,income_middle_dummy_2015)| -.0545097 .0650997 -0.84 0.406 -.1851418 .0761223
cov(black_dummy_2015,income_highest_dummy_2015)| -.1398925 .0596025 -2.35 0.023 -.2594938 -.0202913
cov(black_dummy_2015,education_nodegree_dummy_2015)| .0581537 .0560732 1.04 0.304 -.0543654 .1706728
cov(black_dummy_2015,education_degree_dummy_2015)| -.1593424 .0691944 -2.30 0.025 -.2981912 -.0204936
cov(others_dummy_2015,burden_con_2015)| .0797901 .0993774 0.80 0.426 -.1196252 .2792055
cov(others_dummy_2015,pp_sumscore_2015)| -.1247267 .0887315 -1.41 0.166 -.3027796 .0533261
cov(others_dummy_2015,ss_sumscore_2015)| -.035729 .0967696 -0.37 0.713 -.2299114 .1584534
cov(others_dummy_2015,overall_caregiver_support_2015)| -.0673187 .102811 -0.65 0.515 -.2736242 .1389867
cov(others_dummy_2015,gad2_cont_2017_new)| -.1657358 .053944 -3.07 0.003 -.2739823 -.0574894
cov(others_dummy_2015,age_2015)| -.1552277 .0806392 -1.92 0.060 -.317042 .0065867
cov(others_dummy_2015,gender_2015)| .0549183 .0876185 0.63 0.534 -.120901 .2307377
cov(others_dummy_2015,income_middle_dummy_2015)| -.060682 .083535 -0.73 0.471 -.2283072 .1069432
cov(others_dummy_2015,income_highest_dummy_2015)| -.067295 .0919531 -0.73 0.468 -.2518124 .1172224
cov(others_dummy_2015,education_nodegree_dummy_2015)| -.0326265 .072867 -0.45 0.656 -.1788448 .1135918
cov(others_dummy_2015,education_degree_dummy_2015)| .1237474 .1123541 1.10 0.276 -.1017077 .3492025
cov(burden_con_2015,pp_sumscore_2015)| -.126503 .0744729 -1.70 0.095 -.2759438 .0229378
cov(burden_con_2015,ss_sumscore_2015)| .20376 .0970496 2.10 0.041 .0090158 .3985042
cov(burden_con_2015,overall_caregiver_support_2015)| .0965205 .1091871 0.88 0.381 -.1225794 .3156205
cov(burden_con_2015,gad2_cont_2017_new)| .1958 .0644432 3.04 0.004 .0664854 .3251147
cov(burden_con_2015,age_2015)| -.0381358 .0640562 -0.60 0.554 -.1666741 .0904024
cov(burden_con_2015,gender_2015)| .2020486 .0474854 4.25 0.000 .1067622 .297335
cov(burden_con_2015,income_middle_dummy_2015)| -.0383833 .0649882 -0.59 0.557 -.1687916 .092025
cov(burden_con_2015,income_highest_dummy_2015)| -.006801 .0652831 -0.10 0.917 -.1378012 .1241992
cov(burden_con_2015,education_nodegree_dummy_2015)| .2010744 .06941 2.90 0.006 .061793 .3403558
cov(burden_con_2015,education_degree_dummy_2015)| .0311916 .0835083 0.37 0.710 -.13638 .1987632
cov(pp_sumscore_2015,ss_sumscore_2015)| .2213706 .0584951 3.78 0.000 .1039917 .3387496
cov(pp_sumscore_2015,overall_caregiver_support_2015)| .2787945 .0636334 4.38 0.000 .1511047 .4064843
cov(pp_sumscore_2015,gad2_cont_2017_new)| -.039442 .0651151 -0.61 0.547 -.170105 .091221
cov(pp_sumscore_2015,age_2015)| -.1182591 .0736004 -1.61 0.114 -.2659491 .0294309
cov(pp_sumscore_2015,gender_2015)| -.0255948 .0583083 -0.44 0.663 -.1425989 .0914094
cov(pp_sumscore_2015,income_middle_dummy_2015)| .0562742 .073771 0.76 0.449 -.0917581 .2043064
cov(pp_sumscore_2015,income_highest_dummy_2015)| .3520238 .0650717 5.41 0.000 .221448 .4825996
cov(pp_sumscore_2015,education_nodegree_dummy_2015)| -.005251 .0915988 -0.06 0.955 -.1890575 .1785555
cov(pp_sumscore_2015,education_degree_dummy_2015)| .2114563 .0919304 2.30 0.025 .0269845 .3959281
cov(ss_sumscore_2015,overall_caregiver_support_2015)| .7286871 .0341241 21.35 0.000 .6602121 .7971622
cov(ss_sumscore_2015,gad2_cont_2017_new)| .070241 .072094 0.97 0.334 -.0744263 .2149083
cov(ss_sumscore_2015,age_2015)| -.0880865 .0726559 -1.21 0.231 -.2338812 .0577082
cov(ss_sumscore_2015,gender_2015)| -.0208017 .0605459 -0.34 0.733 -.1422961 .1006926
cov(ss_sumscore_2015,income_middle_dummy_2015)| -.0317639 .0597881 -0.53 0.597 -.1517375 .0882097
cov(ss_sumscore_2015,income_highest_dummy_2015)| .1711175 .0582504 2.94 0.005 .0542296 .2880054
cov(ss_sumscore_2015,education_nodegree_dummy_2015)| -.0420811 .060744 -0.69 0.492 -.1639729 .0798107
cov(ss_sumscore_2015,education_degree_dummy_2015)| .1768812 .0816779 2.17 0.035 .0129824 .34078
cov(overall_caregiver_support_2015,gad2_cont_2017_new)| .1104752 .0808636 1.37 0.178 -.0517896 .2727399
cov(overall_caregiver_support_2015,age_2015)| -.1764987 .0869611 -2.03 0.048 -.3509989 -.0019985
cov(overall_caregiver_support_2015,gender_2015)| -.1115753 .0605257 -1.84 0.071 -.233029 .0098784
cov(overall_caregiver_support_2015,income_middle_dummy_2015)| -.0430981 .0693295 -0.62 0.537 -.1822179 .0960218
cov(overall_caregiver_support_2015,income_highest_dummy_2015)| .2019868 .0465227 4.34 0.000 .1086322 .2953414
cov(overall_caregiver_support_2015,education_nodegree_dummy_2015)| -.0289982 .0655182 -0.44 0.660 -.1604701 .1024737
cov(overall_caregiver_support_2015,education_degree_dummy_2015)| .1504074 .0916671 1.64 0.107 -.0335361 .3343508
cov(gad2_cont_2017_new,age_2015)| .0533754 .0584003 0.91 0.365 -.0638134 .1705643
cov(gad2_cont_2017_new,gender_2015)| .0673104 .0650918 1.03 0.306 -.0633058 .1979266
cov(gad2_cont_2017_new,income_middle_dummy_2015)| .0156109 .0631985 0.25 0.806 -.1112062 .142428
cov(gad2_cont_2017_new,income_highest_dummy_2015)| -.0855794 .0545082 -1.57 0.122 -.1949582 .0237993
cov(gad2_cont_2017_new,education_nodegree_dummy_2015)| -.0389021 .0632003 -0.62 0.541 -.1657228 .0879186
cov(gad2_cont_2017_new,education_degree_dummy_2015)| -.1231722 .0621899 -1.98 0.053 -.2479654 .0016209
cov(age_2015,gender_2015)| -.0073641 .0607158 -0.12 0.904 -.1291993 .1144711
cov(age_2015,income_middle_dummy_2015)| -.0747732 .0739356 -1.01 0.317 -.2231359 .0735896
cov(age_2015,income_highest_dummy_2015)| .0447633 .0718098 0.62 0.536 -.0993335 .1888602
cov(age_2015,education_nodegree_dummy_2015)| -.0734998 .0797145 -0.92 0.361 -.2334587 .0864591
cov(age_2015,education_degree_dummy_2015)| -.1114566 .0829494 -1.34 0.185 -.2779068 .0549936
cov(gender_2015,income_middle_dummy_2015)| -.0492902 .0757511 -0.65 0.518 -.2012958 .1027154
cov(gender_2015,income_highest_dummy_2015)| -.0094492 .0878228 -0.11 0.915 -.1856786 .1667802
cov(gender_2015,education_nodegree_dummy_2015)| .0541409 .0658622 0.82 0.415 -.0780213 .1863032
cov(gender_2015,education_degree_dummy_2015)| .0093998 .0659914 0.14 0.887 -.1230216 .1418213
cov(income_middle_dummy_2015,income_highest_dummy_2015)| -.4175598 .0478969 -8.72 0.000 -.513672 -.3214475
cov(income_middle_dummy_2015,education_nodegree_dummy_2015)| .1096786 .083194 1.32 0.193 -.0572624 .2766196
cov(income_middle_dummy_2015,education_degree_dummy_2015)| -.0362842 .0907088 -0.40 0.691 -.2183048 .1457364
cov(income_highest_dummy_2015,education_nodegree_dummy_2015)| -.0546737 .0745125 -0.73 0.466 -.2041939 .0948465
cov(income_highest_dummy_2015,education_degree_dummy_2015)| .3685904 .0895839 4.11 0.000 .1888272 .5483536
cov(education_nodegree_dummy_2015,education_degree_dummy_2015)| -.4070101 .0374472 -10.87 0.000 -.4821535 -.3318667
----------------------------------------------------------------------------------------------------------------------------------
I would like to seek your insights on whether my approach is appropriate. If it is, how should I interpret the correlations between two dummy variables, or between a dummy variable and a continuous variable?
If this approach is not suitable, could you kindly share any alternative methods for deriving a correlation matrix that would be appropriate for survey-weighted data and datasets including both continuous and categorical variables?
Thank you very much for your time and guidance.
Comment