JTAM (Jurnal Teori dan Aplikasi Matematika) http://journal.ummat.ac.id/index.php/jtam p-ISSN 2597-7512 | e-ISSN 2614-1175 Vol. 8, No. 1, January 2024, pp. 195-205 195 Nonlinear Principal Component Analysis with Mixed Data Formative Indicator Models in Path Analysis Rindu Hardianti 1 , Solimun 2 , Nurjannah 3 , Rosita Hamdan 4 1,2,3 Departement of Statistics, University of Brawijaya, Indonesia 4 Department of Development Economics, University Malaysia Serawak, Malaysia rinduhardianti@student.ub.ac.id 1 , solimun@ub.ac.id 2 , nj_anna@ub.ac.id 3 , hrosita@unimas.my 4 ABSTRACT Article History: Received : 31-08-2023 Revised : 09-12-2023 Accepted : 17-12-2023 Online : 19-01-2024 This research aims to obtain the main component score of the latent variable ability to pay, determine the strongest indicators forming the ability to pay on a mixed scale based on predetermined indicators, and model the ability to pay on time as mediated by fear of paying using path analysis. The data used is secondary data obtained through distributing questionnaires with a mixed data scale. The sampling technique used in the research was purposive sampling. The number of samples used in the research was 100 customers. The method used is nonlinear principal component analysis with path analysis modeling. The results of this research show that of the five indicators formed by the Principal Component, 74.8% of diversity or information is able to be stored, while 25.20% of diversity or other information is not stored (wasted). Credit term is the strongest indicator that forms the ability to pay variable. The variable ability to pay mortgage has a significant effect on payments by mediating the fear of being late in paying with a coefficient of determination of 73.63%. Keywords: Nonlinear Principal Component Analysis; Path Analysis; Mixed Data; Formative Indicator Models. https://doi.org/10.31764/jtam.v8i1.17559 This is an open access article under the CC–BY-SA license ——————————  —————————— A. INTRODUCTION Multivariate analysis is one of statistical analysis that simultaneously analyze several variables in individuals or objects (Astutik et al., 2018). With multivariate analysis, the effect of several variables on other variables can be analyzed at once. Meanwhile, according to Solimun & Fernandes (2017), multivariate analysis can be said to be the use of statistical methods related to several variables where the measurements are carried out from each research object. The variable itself is a characteristic of the subject or object that is relevant to the problem being studied, where there are various variables viewed from various points of view. Based on the measurement process, variables are divided into manifest variables (observable) and latent variables (unobservable). According to Solimun et al. (2017), in general latent variables are defined as variables that cannot be measured directly, but must be through indicators that reflect or structure them. Latent variables can be sorted into variables in the form of psychological attributes such as satisfaction in the form of conception variables, and can also be sorted into latent variables which are factual in nature such as the ability to pay mortgage variables which will be examined