* -------------------------------------------------------------- * Evaluating push-to-web methodology for mixed-mode surveys using address-based samples * * Peter Lynn, University of Essex * * Published in Survey Research Methods * -------------------------------------------------------------- use "\\....\k_hhsamp_ip.dta", clear *** 0. Deriving new hosuehold-level variables ge fullresp=0 replace fullresp=1 if k_ivfho==10 ge anyresp=0 replace anyresp=1 if k_ivfho>=10 & k_ivfho<=16 ge followup=0 replace followup=1 if k_ff_invitew11>=3 ge others=0 replace others=1 if k_ff_invitew11==1|k_ff_invitew11==3 ge anyweb=k_hhmodes recode anyweb (4=2) ge k_modeall=k_ff_gridmodew11 recode k_modeall (3=0) merge 1:1 k_hidp using "\\....\k_hhresp_ip.dta", keepusing(k_hhsize k_dwltype k_housebun k_cduse12 k_pcnet k_pdepa1 k_pdepb1 k_pdepc1 k_ncars k_agechy_dv k_nemp_dv k_npens_dv k_nkids_dv k_nonepar_dv k_hhtype_dv k_tenure_dv k_urban_dv k_fihhmngrs_dv) ge k_numads=k_hhsize-k_nkids_dv ge onead=k_numads recode onead (3/8=2) ge fourads=k_numads recode fourads (8=4) (2/3=1) ge twoads=k_numads recode twoads (3/8=1) ge threeads=k_numads recode threeads (2=1) (4/8=1) drop _merge ge elig=0 replace elig=1 if (k_ivfho<81|k_ivfho>91) & k_outcome_tns>=20 ge id=1 ge k_f2fmode=k_hhmodes recode k_f2fmode (4=2) ge k_webmode=k_hhmodes recode k_webmode (1=4) ge k_twomode=k_hhmodes recode k_twomode (2=1) * Note: there is some further variable derivation in the sample composition section below, consisting solely of * categorisation (or further categorisation) of existing variables at individual level **** 1. Household-Level Analyses *** 1.1 Sample accounting (Table 1)*** ta k_ff_gridmodew11 elig if k_hhorig==18 ta k_ff_invitew11 elig if k_hhorig==18 *** 1.2 Modes experiment *** * Table 2, first panel: svyset k_psu [pweight=id], strata(k_strata) singleunit(centered) svy: ta anyresp k_ff_gridmodew11 if (k_ivfho<81|k_ivfho>91) & k_outcome_tns>=20 & k_hhorig==18, col svy: ta fullresp k_ff_gridmodew11 if (k_ivfho<81|k_ivfho>91) & k_outcome_tns>=20 & k_hhorig==18, col * Table 2, second panel, proportions: svy: ta k_hhmodes k_ff_gridmodew11 if (k_ivfho<81|k_ivfho>91) & k_outcome_tns>=20 & k_hhorig==18 & k_hhmodes>0, col * Table 2, second panel, P-values: svy: ta k_f2fmode k_ff_gridmodew11 if (k_ivfho<81|k_ivfho>91) & k_outcome_tns>=20 & k_hhorig==18 & k_hhmodes>0, col svy: ta k_webmode k_ff_gridmodew11 if (k_ivfho<81|k_ivfho>91) & k_outcome_tns>=20 & k_hhorig==18 & k_hhmodes>0, col svy: ta k_twomode k_ff_gridmodew11 if (k_ivfho<81|k_ivfho>91) & k_outcome_tns>=20 & k_hhorig==18 & k_hhmodes>0, col *** 1.3 Experiment: Inviting all household members *** svyset k_psu [pweight=id], strata(k_strata) singleunit(centered) * Table 4, top panel: svy: ta anyresp others if k_ff_invitew11>0 & (k_ivfho<81|k_ivfho>91) & k_outcome_tns>=20 & k_hhorig==18, col svy: ta fullresp others if k_ff_invitew11>0 & (k_ivfho<81|k_ivfho>91) & k_outcome_tns>=20 & k_hhorig==18, col * Table 4, lower panel: svy: ta k_numads others if k_ff_invitew11>0 & (k_ivfho<81|k_ivfho>91) & k_outcome_tns>=20 & k_hhorig==18 & k_hhmodes>0, col svy: ta onead others if k_ff_invitew11>0 & (k_ivfho<81|k_ivfho>91) & k_outcome_tns>=20 & k_hhorig==18 & k_hhmodes>0, col svy: ta twoads others if k_ff_invitew11>0 & (k_ivfho<81|k_ivfho>91) & k_outcome_tns>=20 & k_hhorig==18 & k_hhmodes>0, col svy: ta threeads others if k_ff_invitew11>0 & (k_ivfho<81|k_ivfho>91) & k_outcome_tns>=20 & k_hhorig==18 & k_hhmodes>0, col svy: ta fourads others if k_ff_invitew11>0 & (k_ivfho<81|k_ivfho>91) & k_outcome_tns>=20 & k_hhorig==18 & k_hhmodes>0, col *** 1.4 Experiment: Introducing the CAPI follow-up * Table 5, top panel: svy: ta anyresp followup if k_ff_invitew11>0 & (k_ivfho<81|k_ivfho>91) & k_outcome_tns>=20 & k_hhorig==18, col svy: ta fullresp followup if k_ff_invitew11>0 & (k_ivfho<81|k_ivfho>91) & k_outcome_tns>=20 & k_hhorig==18, col * Table 5, lower panel: svy: ta anyweb followup if k_ff_invitew11>0 & (k_ivfho<81|k_ivfho>91) & k_outcome_tns>=20 & k_hhorig==18 & k_hhmodes>0, col **** 2. Individual-Level Analyses *** 2.1 Modes experiment * Table 2, third panel: merge 1:m k_hidp using "\\....\k_indresp_ip.dta" svy: ta k_indmode k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2), col ge k_proxy=k_indmode recode k_proxy (3=1) ge k_capi=k_indmode recode k_capi (-7=3) ge k_cawi=k_indmode recode k_cawi (-7=4) (1=4) svy: ta k_proxy k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2), col svy: ta k_capi k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2), col svy: ta k_cawi k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2), col ** Sample Composition ** * Step 1: Bivariate tests svy: ta k_hhsize k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2), col ge k_hhsize3=k_hhsize recode k_hhsize3 (4/12=3) svy: ta k_hhsize3 k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2), col ge k_hhsize4=k_hhsize recode k_hhsize4 (3/12=2) svy: ta k_hhsize4 k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2), col ge k_hhsize5=k_hhsize recode k_hhsize5 (4/12=3) (2=1) svy: ta k_hhsize5 k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2), col svy: ta k_sex k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2), col svy: ta k_agegr10_dv k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) & k_agegr10_dv>0, col svy: ta k_pcnet k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2), col svy: ta k_dwltype k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2), col ge k_dwltype2=k_dwltype recode k_dwltype2 (3/8=2) svy: ta k_dwltype2 k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2), col svy: ta k_ncars k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2), col ge k_ncars2=k_ncars recode k_ncars2 (4/13=3) svy: ta k_ncars2 k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) & k_ncars2>=0, col svy: ta k_nkids_dv k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2), col ge k_nkids2=k_nkids_dv recode k_nkids2 (2/5=1) svy: ta k_nkids2 k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2), col ge k_nkids3=k_nkids_dv recode k_nkids3 (5=4) svy: ta k_nkids3 k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2), col svy: ta k_hhtype_dv k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2), col ge k_onead=k_hhtype_dv recode k_onead (4/5=3) (2=1) (6/23=1) svy: ta k_onead k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2), col ge k_adults=k_hhtype_dv recode k_adults (4/5=3) (2=1) (6/23=0) svy: ta k_adults k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2), col svy: ta k_cduse12 k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) & k_cduse12>=0, col svy: ta k_tenure_dv k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) & k_tenure_dv>=0, col ge k_tenurea=k_tenure_dv recode k_tenurea (6/7=5) (4/8=3) svy: ta k_tenurea k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) & k_tenure_dv>=0, col svy: ta k_npens_dv k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2), col svy: ta k_urban_dv k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2), col svy: regress k_fihhmngrs_dv k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) ge k_hiinc=0 replace k_hiinc=1 if k_fihhmngrs_dv>=8333.3 svy: ta k_hiinc k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2), col ge k_vhiinc=0 replace k_vhiinc=1 if k_fihhmngrs_dv>=10000 svy: ta k_vhiinc k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2), col sum k_fihhmngrs_dv if k_hhorig==18 & (k_ivfio==1|k_ivfio==2), detail ge percentile05=r(p5) ta percentile05 k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) ge k_lowinc=0 replace k_lowinc=1 if k_fihhmngrs_dv<=799.7 svy: ta k_lowinc k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2), col svy: ta k_nemp_dv k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2), col ge k_ukborn2=k_ukborn recode k_ukborn2 (2/4=1) svy: ta k_ukborn2 k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2), col ge k_hiqual2=k_hiqual_dv recode k_hiqual2 (-9=9) svy: ta k_hiqual2 k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2), col ge k_health2=k_health recode k_health2 (-1=2) (-2=1) svy: ta k_health2 k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) & k_health>0, col svy: ta k_hl2gp k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) & k_hl2gp>=0, col * Note: not asked for proxies svy: ta k_smoker k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) & k_smoker>=0, col * Note: not asked for proxies ge k_aidhh2=k_aidhh recode k_aidhh2 (-8=2) (-1=2) svy: ta k_aidhh2 k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) & k_smoker>=0, col svy: ta k_jbhas k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) & k_jbhas>=0, col ge k_jbhas2=k_jbhas recode k_jbhas2 (-1=2) svy: ta k_benbase96 k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) & k_benbase96>=0, col * Note: not asked for proxies svy: ta k_marstat k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) & k_marstat>=0, col * Note: not asked for proxies ge k_marstat2=k_marstat recode k_marstat2 (3=2) (5/6=4) svy: ta k_marstat2 k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) & k_marstat>=0, col svy: ta k_employ k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) & k_employ>=0, col * Note: not asked for proxies svy: regress k_scghq1_dv k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) & k_scghq1_dv>0 svy: regress k_sf12pcs_dv k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) & k_scghq1_dv>0 svy: regress k_sf12mcs_dv k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) & k_scghq1_dv>0 ge k_racel2=k_racel_dv recode k_racel2 (2=1) (5/17=4) svy: ta k_racel2 k_ff_gridmodew11 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) & k_racel2>=0, col * Note: not asked for proxies * Step 2: Logit model (forward stepwise) - Table 3 svy: logit k_modeall k_sex if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.23, drop k_sex svy: logit k_modeall i.k_hhsize3 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.068 for k_hhsize3==3 svy: logit k_modeall i.k_agegr10_dv if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) & k_agegr10_dv>=0 ge k_agegr10_2=k_agegr10_dv recode k_agegr10_2 (-9=.) svy: logit k_modeall i.k_agegr10_2 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * all P>=0.5, drop k_agegr10_2 ge k_pcnet2=k_pcnet recode k_pcnet2 (-1=1) svy: logit k_modeall k_pcnet2 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.79, drop svy: logit k_modeall k_dwltype2 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.84, drop ge k_ncars3=k_ncars2 recode k_ncars3 (-1=.) (-2=.) svy: logit k_modeall i.k_ncars3 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.075 for ncars==2, drop svy: logit k_modeall i.k_nkids3 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.0.016 for nkids==4 (i.e. 4+), keep svy: logit k_modeall i.k_nkids3 i.k_adults if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.005, keep ge k_cduse122=k_cduse12 recode k_cduse122 (-1=.) (-2=.) svy: logit k_modeall i.k_nkids3 i.k_adults k_cduse122 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.43, drop ge k_tenureb=k_tenurea recode k_tenureb (-9=.) svy: logit k_modeall i.k_nkids3 i.k_adults i.k_tenureb if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P>=0.12, drop svy: logit k_modeall i.k_nkids3 i.k_adults i.k_npens_dv if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.88, 0.16, drop svy: logit k_modeall i.k_nkids3 i.k_adults k_urban_dv if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.15, drop svy: logit k_modeall i.k_nkids3 i.k_adults k_hiinc if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.48, drop svy: logit k_modeall i.k_nkids3 i.k_adults k_lowinc if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.10, drop svy: logit k_modeall i.k_nkids3 i.k_adults i.k_nemp_dv if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * all P>0.25, drop ge k_ukborn3=k_ukborn2 recode k_ukborn3 (-1=1) svy: logit k_modeall i.k_nkids3 i.k_adults k_ukborn3 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.95, drop svy: logit k_modeall i.k_nkids3 i.k_adults i.k_hiqual2 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * all P>=0.14, drop svy: logit k_modeall i.k_nkids3 i.k_adults k_health2 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.51, drop svy: logit k_modeall i.k_nkids3 i.k_adults k_hl2gp if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.29, drop svy: logit k_modeall i.k_nkids3 i.k_adults k_smoker if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.39, drop svy: logit k_modeall i.k_nkids3 i.k_adults k_aidhh2 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.20, drop svy: logit k_modeall i.k_nkids3 i.k_adults k_jbhas2 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.92, drop ge k_benbase962=k_benbase96 recode k_benbase962 (-8/-1=.) svy: logit k_modeall i.k_nkids3 i.k_adults k_benbase962 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.76, drop ge k_marstat3=k_marstat2 recode k_marstat3 (-8/-1=.) svy: logit k_modeall i.k_nkids3 i.k_adults k_marstat3 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.22, drop ge k_racel3=k_racel2 recode k_racel3 (-9=.) svy: logit k_modeall i.k_nkids3 i.k_adults k_racel3 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.88, drop ge k_scghq1_2=k_scghq1_dv recode k_scghq1_2 (-9/-7=.) svy: logit k_modeall i.k_nkids3 i.k_adults k_scghq1_2 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.32, drop ge k_sf12pcs_2=k_sf12pcs_dv recode k_sf12pcs_2 (-9/-7=.) svy: logit k_modeall i.k_nkids3 i.k_adults k_sf12pcs_2 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.54, drop ge k_sf12mcs_2=k_sf12mcs_dv recode k_sf12mcs_2 (-9/-7=.) svy: logit k_modeall i.k_nkids3 i.k_adults k_sf12mcs_2 if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.84, drop * Final model: svy: logit k_modeall i.k_nkids3 i.k_adults if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) svy: logistic k_modeall i.k_nkids3 i.k_adults if k_hhorig==18 & (k_ivfio==1|k_ivfio==2) /* Survey: Logistic regression Number of strata = 60 Number of obs = 798 Number of PSUs = 117 Population size = 798 Design df = 57 F( 6, 52) = 2.92 Prob > F = 0.0157 ------------------------------------------------------------------------------ | Linearized k_modeall | Odds Ratio Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- k_nkids3 | 1 | 1.055606 .2756346 0.21 0.837 .6257782 1.780668 2 | 1.605628 .5001219 1.52 0.134 .8605235 2.995898 3 | 1.885752 1.699035 0.70 0.484 .3104084 11.45607 4 | .1587293 .1253628 -2.33 0.023 .0326444 .7718016 | k_adults | 1 | 1.053969 .3250271 0.17 0.865 .5683802 1.954416 3 | .4880608 .1212153 -2.89 0.005 .296814 .8025341 | _cons | 3.062018 .4790258 7.15 0.000 2.2385 4.188498 ------------------------------------------------------------------------------ Note: _cons estimates baseline odds. Note: Strata with single sampling unit centered at overall mean. */ *** 2.2 Who responds online? * Model of mode of response - Table 6 ge k_online=9 replace k_online=1 if k_cawi==3 replace k_online=0 if k_cawi==4 fre k_online if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) svy: logit k_online k_sex if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.92, drop svy: logit k_online i.k_hhsize7 if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.049 for 2/3 person hhds; 0.09 for 5+. Both less likely to respond online than 1-person ge k_hhsize8=k_hhsize7 recode k_hhsize8 (2=1) (4=1) svy: logit k_online i.k_hhsize8 if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.13 ge k_hhsize9=k_hhsize7 recode k_hhsize9 (4=2) (5=2) svy: logit k_online i.k_hhsize9 if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.15, drop svy: logit k_online i.k_agegr10_2 if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) ge k_agegr10_3=k_agegr10_2 recode k_agegr10_3 (3/6=2) (8=7) svy: logit k_online i.k_agegr10_3 if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.04 // age 60+ less likely, keep ge k_pcnet3=k_pcnet recode k_pcnet3 (-1=2) (.=1) (-2=1) svy: logit k_online i.k_agegr10_3 i.k_pcnet3 if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.07 // no internet less likely, drop recode k_dwltype2 (-9=1) (-8=1) (.=1) svy: logit k_online i.k_agegr10_3 i.k_dwltype2 if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.33, drop svy: logit k_online i.k_agegr10_3 i.k_nkids_dv if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.005 for n=1 ge k_nkids4=k_nkids_dv recode k_nkids4 (3/5=2) svy: logit k_online i.k_agegr10_3 i.k_nkids4 if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.005 for n=1 (less likely); n.s. (0.33) for 2+ svy: logit k_online i.k_agegr10_3 i.k_nkids4 k_jbhas2 if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.45, drop jbhas svy: logit k_online i.k_agegr10_3 i.k_nkids4 k_ncars3 if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.15, drop ncars svy: logit k_online i.k_agegr10_3 i.k_nkids4 k_cduse122 if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.12, drop cduse svy: logit k_online i.k_agegr10_3 i.k_nkids4 k_tenure4 if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.41, drop tenure svy: logit k_online i.k_agegr10_3 i.k_nkids4 k_npens_dv if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.051 and agegr10_3 becomes n.s. (P=0.70), drop npens; keep agegr10 svy: logit k_online i.k_agegr10_3 i.k_nkids4 k_ukborn3 if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.19, drop ukborn svy: logit k_online i.k_agegr10_3 i.k_nkids4 k_urban_dv if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.86, drop urban svy: logit k_online i.k_agegr10_3 i.k_nkids4 k_hiinc if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.13, drop hiinc svy: logit k_online i.k_agegr10_3 i.k_nkids4 k_lowinc if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.89, drop lowinc svy: logit k_online i.k_agegr10_3 i.k_nkids4 k_nemp_dv if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.15, drop nemp svy: logit k_online i.k_agegr10_3 i.k_nkids4 i.k_hiqual2 if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) ge k_hiqual3=k_hiqual2 recode k_hiqual3 (5=4) (9=4) (2/3=1) ge k_hiqual4=k_hiqual2 recode k_hiqual4 (3/9=2) svy: logit k_online i.k_agegr10_3 i.k_nkids4 i.k_hiqual3 if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.01 - less likely if less than A levels svy: logit k_online i.k_agegr10_3 i.k_nkids4 i.k_hiqual4 if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.02 - less likely if less than a degree, keep hiqual4 svy: logit k_online i.k_agegr10_3 i.k_nkids4 i.k_hiqual4 k_health2 if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.03 - more likely, keep health svy: logit k_online i.k_agegr10_3 i.k_nkids4 i.k_hiqual4 k_health2 k_hl2gp if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.24, drop h12gp svy: logit k_online i.k_agegr10_3 i.k_nkids4 i.k_hiqual4 k_health2 k_smoker if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.042; but k_health becomes n.s. (0.052, was 0.031) - choose health svy: logit k_online i.k_agegr10_3 i.k_nkids4 i.k_hiqual4 k_health2 k_aidhh2 if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * omitted svy: logit k_online i.k_agegr10_3 i.k_nkids4 i.k_hiqual4 k_health2 k_marstat3 if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.86, drop marstat svy: logit k_online i.k_agegr10_3 i.k_nkids4 i.k_hiqual4 k_health2 k_racel3 if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.26, drop racel svy: logit k_online i.k_agegr10_3 i.k_nkids4 i.k_hiqual4 k_health2 k_scghq1_2 if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.89, drop scghq svy: logit k_online i.k_agegr10_3 i.k_nkids4 i.k_hiqual4 k_health2 k_sf12pcs_2 if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.39, drop sf12pcs svy: logit k_online i.k_agegr10_3 i.k_nkids4 i.k_hiqual4 k_health2 k_sf12mcs_2 if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) * P=0.16, drop sf12mcs **FINAL MODEL: svy: logit k_online i.k_agegr10_3 i.k_nkids4 i.k_hiqual4 k_health2 if k_ff_gridmodew11==3 & k_hhorig==18 & (k_ivfio==1|k_ivfio==2) /* Survey: Logistic regression Number of strata = 53 Number of obs = 206 Number of PSUs = 80 Population size = 206 Design df = 27 F( 5, 23) = 6.02 Prob > F = 0.0011 ------------------------------------------------------------------------------- | Linearized k_online | Coef. Std. Err. t P>|t| [95% Conf. Interval] --------------+---------------------------------------------------------------- 7.k_agegr10_3 | -1.470233 .4401169 -3.34 0.002 -2.373278 -.5671878 | k_nkids4 | 1 | -2.621604 .8251792 -3.18 0.004 -4.314732 -.9284766 2 | -.7189816 .525952 -1.37 0.183 -1.798146 .3601827 | 2.k_hiqual4 | -.9837185 .4121934 -2.39 0.024 -1.82947 -.1379674 k_health2 | .7660139 .3365381 2.28 0.031 .0754948 1.456533 _cons | -.5231277 .6879468 -0.76 0.454 -1.934678 .8884225 ------------------------------------------------------------------------------- Note: Strata with single sampling unit centered at overall mean. */ *** 3 Field efforts - reported in text of section 4.1 use "\\....\k_callrec_ip.dta", clear sort k_hidp keep if k_hidp~=k_hidp[_n-1]|_n==1 merge 1:1 k_hidp using "\\....\k_hhsamp_ip.dta" ge nc=0 replace nc=1 if k_outcome_tns==20|k_outcome_tns==21|k_outcome_tns==40|k_outcome_tns==82|k_outcome_tns==89|k_outcome_tns==110|k_outcome_tns==205 mean k_ivtnc if k_ff_gridmodew11==1 & k_hhorig==18 & nc==1 mean k_ivtnc if k_ff_gridmodew11==3 & k_hhorig==18 & nc==1