To whom it may concern,
I would like to ask for some info about the weights of SC2. If I consider only wave 3 (grade 1, augmented sample so without considering the students in wave 1 and 2 who made it to wave 3), and exploit the info given by targets and parents, 1) should I use „w_tp3“?
2) Is this the weight called „adjusted“ in the technical reports (e.g., the Supplement of Technical report: Weighting the Kindergarten Cohort of the National Educational Panel Study, Panel Cohort by Aßmann et al., 2012) of starting cohort 2? Put differently, is this the weight that adjusts for unit-nonresponse?
3) According to the technical report WP_LXVI, I would like to use the nonresponse adjustment at the level of students taking part in Wave 3 ( stage iii on page 7). However, this weight (and the other adjustments for unit-nonresponse) accounts for: federal state, gender, month and year of birth, nationality, when the student started schooling, special educational needs, and German as native language. Is it correct to assume that they do not account for social background/socioeconomic status?
Vielen Dank im Voraus für die Hilfe
Claudia
Dear Claudia,
thank you for your questions. I will give some clarification.
-
Given your focus is on the comparison of target and parent information in Wave 3, „w_tp3“ from WeightsElementarySchool seems the correct weight to use. However, you additionally restrict yourself to the augmentation sample (Group 1) and „w_tp3“ is the cross-sectional weight for Group 1 and Group 3, i.e. for all target children in Elementary School. There were no weights exclusively for Group 1 provided.
-
Cross-sectional weights include correction for the sampling design (i.e., w_t) and are additionally adjusted for nonresponse (i), (ii), and (iii) in WP_LXVI. The nonresponse correction (iii) at the level of students (or students and parents jointly) is repeated for each follow-up wave. Proceeding this way, one gets the so called nonresponse adjusted design-weights. The cross-sectional weights correct for wave-nonresponse and are given for participants only. Please differentiate carefully between „w_t3“ and „w_tp3“, „w_t3“ weights all participating target children, opposed to „w_tp3“ weighting the intersection of participating target children and parents. You can confirm yourself by looking at the sum of the weights (sum of „w_t3“ = 6,733 versus sum of „w_tp3“ = 5,636).
-
The predictors you have listed were those being available, not necessarily being used. Furthermore, there were different models for children getting panel members (Table 4, in WP_LXVI), target children participating (Table 2), and target children and parents jointly participating (Table 3). The predictors in Table 2 and 3 are selected due to significance, i.e. significantly influence the willingness to participate. In the nonresponse model of Group 1 used to derive the adjustment factors for target participation only (multilevel probit model with a random intercept at the school level), the following predictors were included: native language, special educational needs and a random intercept on the school level were included, see Table 2 in WP_LXVI. In the nonresponse model of Group 1 used to derive the adjustment factors for joint participation of target child and parent (bivariate probit model), again native language and special educational needs were finally selected, see Table 3 in WP_LXVI. At time of sampling no information on social background/ socioeconomic status (of parents) was available for Group 1, so could not be considered for nonresponse adjustment.
Kind regards,
Ariane