## Handling missing data: applications to environmental analysisDesigned for use as a practical guide, this volume features a wide range of techniques for analyzing and filling gaps in time series data. It starts with a description of the methods that can be used to replace absent data based on the distribution of the gaps and then classifies the latter according to their length. Various specialized algorithms and techniques are also detailed.Particular attention is paid to "nearest neighbour" techniques that are able to fill gaps of up to eight hours, while Auto Regressive Models and Artificial Neural Networks (ANN) with the ability to fill longer gaps are also explored. ANN's various architectures, features and their application to air quality time series remediation are highlighted.The text also contains an extensive literature review. |

### What people are saying - Write a review

We haven't found any reviews in the usual places.

### Contents

Imputation techniques | 12 |

Data validation and data gaps in environmental | 29 |

Statistical modelling of the remediation | 51 |

Copyright | |

8 other sections not shown

### Common terms and phrases

air pollution algorithm applications approach ARMA Artificial Neural Networks Atmospheric Environment autocorrelation Autoregressive back-propagation behaviour Box-Jenkins complete data component computed considered correlation function data filling data gaps Data Mining data points data remediation data set database datum distribution elementary data environmental equation error function estimate example Figure Gradient hidden layer Hot-Deck Imputation line shows linear interpolation linear regression Maximum Likelihood Mean Imputation measurements meteorological methods missing data missing values monitoring stations Moving Average Multiple Imputation Nearest Neighbour nearest neighbour algorithm neurons nodes non-linear normalised output Ozone time series parameters performance pre-processed network problem procedure random record regression model related observed values replaced residuals sample Series Analysis shows the predicted signal simulation Smooth Fill SO2 concentrations standard deviation stationary stochastic structure techniques Text Mining training epochs training set TRAINLM Transfer Function trend units Validation MSE variables variance vector Voronoi diagram weights wind speed