Quantile Regression

10Quantile Regressions of a Latent Variable

10.1 Introduction

The data used in this chapter is in BBM.wf1, which contains several baby birth measurements—such as BBW (baby birth weight), FUNDUS, and MUAC (mid‐upper arm circumference)—and mother indicators or variables. The data points are listed in Figure 10.1, and include ordinal, dummy, and numerical variables. The data file is provided by Dr. Lilis Heri Miscicih (graduated from the School of Public Health, University of Indonesia, and one of my advisories), and after graduated she is a part‐time lecturer in the School of Public Health.

Similar to all types of quantile regressions (QRs) presented in previous chapters, we can easily develop QRs using each baby birth indicator, namely BBW, FUNDUS, and MUAC, as a dependent variable (DV). So we can have three sets of QRs of each baby birth indicator as a DV having various alternative independent variables (IVs) in BBW.wf1. For instance, we can easily develop a set of QRs of FUNDUS on various selected sets of the categorical or/and numerical variables of the mother indicators.

In this case, we will develop a baby latent variable (BLV) based on the three baby birth indicators, BBW, FUNDUS, and MUAC, and apply various QRs of BLV on selected mother indicators. We will take this approach instead of having three sets of QRs, each for BBW, FUNDUS, and MUAC, since there is no a multivariate QR. In addition, we will also develop a mother latent variable (MLV) based on the two ordinal ...