Modelling Return on Assets (ROA) using nonparametric regression spline truncated for longitudinal data

Nonparametric regression is one of the statistical methods used to see the relationship between response variables and predictor variables with unknown curves. The method often used in estimating the curve of nonparametric regression is the spline method. Spline is one of the nonparametric regression approaches that have very specific and very good statistical and visual interpretations. Spline truncated nonparametric regression is not only used to modelling data with cross-section type but can be used to modelling longitudinal data. Longitudinal data is data whose observations are repeated on each subject within a certain time. In this research, the modelling of Return On Assets (ROA) will be based on the factors that influence it with using 30 subjects of go public banks listed on the Indonesia stock exchange during 2012-2018 using nonparametric regression spline truncated for longitudinal data, with the weighting matrix using the variance-covariance matrix of errors. Based on the analysis data that has been done by using the weighting matrix of the error variance-covariance matrix and one-knot point with 14 Increments, the best model is obtained, namely using the weighting matrix based of the error variance-covariance matrix one-knot point with the smallest GCV value of 15.12 with MSE value of 0.739 and R ² value of 85.56%.