On Fitting a Model to a Population Time Series With Missing Values

Abstract

Existing methods for fitting a population model to time series data typically assume that the time series is complete. When there are missing values, it is common practice to substitute interpolated values. When the proportion of values that are missing is large, this can lead to bias in model-fitting. Here, we describe a maximum likelihood approach that allows explicitly for missing values. The approach is applied to a long weekly time series of the dinoflagellate Peridinium gatunense in Lake Kinneret, Israel, in which around 35% of the values are missing.