The singular spectrum analysis for time series With Missing data

Missing data has been a major problem in a lot of statistical researches such as statistical time series analysis, and usually in case of occurrence leads to wrong conclusion about the main population, and also directs statistical analysis to the skewed results. In this research besides a variety of lost data, different retrieval methods are expressed for these data. Then we introduce analysis of singular eigenvalues as a non-parametric analysis approach in time series analysis. This method breaks down time series into independent components, such as process, harmonic and noise, and requires no statistical hypothesis, such as stability conditions, normality of errors and limitation of the number of observations and easily it can be used in recovery of missing data. Then the efficiency of this method will be evaluated by simulation methods and we analyze the results.