Assessment of procedures for air pollution time-series reconstruction

The time-series are widely used in ecological studies. The results derived from them serve to make better environmental health policy. That´s why the data collected in the series must be very reliable. However, these time-series often have missing values. There are several ways of imputing missing values, but it is very common to produce significant biases in the statistical properties of the original series imputed. It is not easy to assess these statistics alterations, given the non-normality nature of the time-series. In this paper, five different procedures for imputing missing values are presented, and the statistics changes are evaluated using vigintiles and intervigintile ranges.