Forecasting with full use of data without interpolation on logistic curve model with missing data

Abstract

We propose a forecasting method based on a logistic curve model with missing data, which are a ubiquitous problem in social science forecasting, especially in marketing. The method completely recovers parameters of the difference equation when data are on an exact solution curve because it uses an unequal step difference equation that has an exact solution. It makes full use of data without wasting any data or generating plausible data for the missing data. It only requires regression analysis and a simple optimization technique and showed better fitting than two conventional methods for three actual datasets.