Explore chapters and articles related to this topic
Measuring construction activity in the UK
Published in Stephen Gruneberg, Global Construction Data, 2019
The ONS also adjusts for seasonal effects, which are effects that consistently appear in a month or quarter each year, such as Christmas or Summer for example. The purpose of this is to allow researchers to compare construction activity in a certain month or quarter with other months or quarters in the same year. For instance, a considerable proportion of the UK construction industry closes between Christmas and New Year. Consequently, without an adjustment for this, construction output in each December would almost always be lower than in November but this would reveal little about building activity occurring within the sector during this period. If an adjustment is made for a substantial proportion of the construction industry closing down between Christmas and New Year, then it facilitates a comparison of December’s activity with that of November of that year or any month in any year. The seasonal adjustment, using historic trends over time, provides a clearer view of the underlying trends. The ONS uses a seasonal adjustment method called X-13ARIMA-SEATS, which was developed by the United States Census Bureau (US Census Bureau, n.d.) and is recommended for use by Eurostat, the European Commission’s statistical agency (Eurostat, 2015). It should be noted that, as the monthly UK data have only been available since 2010, the seasonal adjustment for each month is currently based upon a small sample in an industry where activity is often volatile. The Eurostat statistical agency recommends that five years is the minimum time span needed to establish a seasonal pattern in monthly data for seasonal adjustment and, by the beginning of 2019, the ONS had almost nine years of monthly data. Nevertheless, significant revisions to seasonal patterns in the construction output data may occur until the seasonal adjustment has been developed in the longer-term.
A novel seasonal decomposition-based short-term forecasting framework with Google Trends data
Published in Yafei Zheng, Kin Keung Lai, Shouyang Wang, Forecasting Air Travel Demand, 2018
Yafei Zheng, Kin Keung Lai, Shouyang Wang
When implementing a short-term air travel demand forecasting with monthly or quarterly historical data, seasonality is one of the most significant data patterns. To achieve a good forecasting performance, the seasonality should be carefully measured and adjusted in order to understand the underlying trends when forecasting the future demand. This purpose raised the importance of the seasonal adjustment. A seasonal adjustment is any method that can remove the seasonal component out of a time series. The main objective of seasonal adjustment can be stated as ‘to simplify the data so that they may be more easily interpreted … without a significant loss of information’ (Bell and Hillmer, 1984). The most popular methods for seasonal adjustment can be classified into two groups, i.e., the parametric (or model-based) methods and the non-parametric methods. The first group of seasonal adjustment methods is usually developed based on parametric models, such as the SARIMA models. This group mainly includes the TRAMO/SEATS method developed by Bank of Spain and the STAMP method developed by Andrew Harvey. The TRAMO/SEATS method, using the ARIMA model as its basis for seasonal adjustment, is widely used worldwide. The second group of seasonal adjustment methods mainly includes the X-11 family developed by the US Census Bureau, and the SABL and STL methods developed by Bell Labs, where the X-11 family is the most popular seasonal adjustment in related literature and also in practice. The latest version of an X-11 method is named the X-13-ARIMA-SEATS method, which combines both X-11 and SEATS modules to provide comprehensive seasonal adjustment functions. Generally, understanding and capturing the seasonal patterns and underlying trend component are extremely important for any empirical economic analysis, including the short-term air travel demand forecasting in this chapter.
Trends in MODIS and AERONET derived aerosol optical thickness over Northern Europe
Published in Tellus B: Chemical and Physical Meteorology, 2019
PAUL GLANTZ, EYAL FREUD, CHRISTER JOHANSSON, KEVIN J. NOONE, MATTHIAS TESCHE
Temporal autocorrelation can be a confounding factor when analysing trends in time series data (Weatherhead et al., 1998; von Storch and Zwiers, 1999). Temporal autocorrelation, if present, would effectively reduce the number of independent observations in a given time series. Autocorrelation has been taken in consideration in the trend analyses by using a stable seasonal adjustment on the current time series of AOT (Brockwell and Davis, 2002). In addition, temporal autocorrelation has been taken in consideration as well by calculating the autocorrelation function (Box et al., 1994) for all cases investigated in the present study. The method described in von Storch and Zwiers (1999) was used to calculate adjusted values for the degrees of freedom in our samples. Using this method, the adjusted degrees of freedom were substantially lower than the number of samples, particularly for the trends estimated based on daily AOTs: more than one order in magnitude. The significance of the trends were tested based on a z-test and the adjusted degrees of freedom values.
Copernicus Marine Service Ocean State Report, Issue 4
Published in Journal of Operational Oceanography, 2020
Karina von Schuckmann, Pierre-Yves Le Traon, Neville Smith, Ananda Pascual, Samuel Djavidnia, Jean-Pierre Gattuso, Marilaure Grégoire, Glenn Nolan, Signe Aaboe, Enrique Álvarez Fanjul, Lotfi Aouf, Roland Aznar, T. H. Badewien, Arno Behrens, Maristella Berta, Laurent Bertino, Jeremy Blackford, Giorgio Bolzon, Federica Borile, Marine Bretagnon, Robert J.W. Brewin, Donata Canu, Paola Cessi, Stefano Ciavatta, Bertrand Chapron, Thi Tuyet Trang Chau, Frédéric Chevallier, Boriana Chtirkova, Stefania Ciliberti, James R. Clark, Emanuela Clementi, Clément Combot, Eric Comerma, Anna Conchon, Giovanni Coppini, Lorenzo Corgnati, Gianpiero Cossarini, Sophie Cravatte, Marta de Alfonso, Clément de Boyer Montégut, Christian De Lera Fernández, Francisco Javier de los Santos, Anna Denvil-Sommer, Álvaro de Pascual Collar, Paulo Alonso Lourenco Dias Nunes, Valeria Di Biagio, Massimiliano Drudi, Owen Embury, Pierpaolo Falco, Odile Fanton d’Andon, Luis Ferrer, David Ford, H. Freund, Manuel García León, Marcos García Sotillo, José María García-Valdecasas, Philippe Garnesson, Gilles Garric, Florent Gasparin, Marion Gehlen, Ana Genua-Olmedo, Gerhard Geyer, Andrea Ghermandi, Simon A. Good, Jérôme Gourrion, Eric Greiner, Annalisa Griffa, Manuel González, Annalisa Griffa, Ismael Hernández-Carrasco, Stéphane Isoard, John J. Kennedy, Susan Kay, Anton Korosov, Kaari Laanemäe, Peter E. Land, Thomas Lavergne, Paolo Lazzari, Jean-François Legeais, Benedicte Lemieux, Bruno Levier, William Llovel, Vladyslav Lyubartsev, Pierre-Yves Le Traon, Vidar S. Lien, Leonardo Lima, Pablo Lorente, Julien Mader, Marcello G. Magaldi, Ilja Maljutenko, Antoine Mangin, Carlo Mantovani, Veselka Marinova, Simona Masina, Elena Mauri, J. Meyerjürgens, Alexandre Mignot, Robert McEwan, Carlos Mejia, Angélique Melet, Milena Menna, Benoît Meyssignac, Alexis Mouche, Baptiste Mourre, Malte Müller, Giulio Notarstefano, Alejandro Orfila, Silvia Pardo, Elisaveta Peneva, Begoña Pérez-Gómez, Coralie Perruche, Monika Peterlin, Pierre-Marie Poulain, Nadia Pinardi, Yves Quilfen, Urmas Raudsepp, Richard Renshaw, Adèle Révelard, Emma Reyes-Reyes, M. Ricker, Pablo Rodríguez-Rubio, Paz Rotllán, Eva Royo Gelabert, Anna Rubio, Inmaculada Ruiz-Parrado, Shubha Sathyendranath, Jun She, Karina von Schuckmann, Cosimo Solidoro, Emil V. Stanev, Joanna Staneva, Andrea Storto, Jian Su, Tayebeh Tajalli Bakhsh, Gavin H. Tilstone, Joaquín Tintoré, Cristina Toledano, Jean Tournadre, Benoit Tranchant, Rivo Uiboupin, Arnaud Valcarcel, Nadezhda Valcheva, Nathalie Verbrugge, Mathieu Vrac, J.-O. Wolff, Enrico Zambianchi, O. Zielinski, Ann-Sofie Zinck, Serena Zunino
The monthly anomalies were aggregated over the full globe and in regions using grid cell area weighted averaging to generate time series. These were further analysed to calculate a linear trend estimate using the procedure described in Mulet et al. (2016). In summary, the time series were each decomposed into a residual seasonal component, a trend component and errors using the X-11 seasonal adjustment method (e.g. Pezzulli et al. 2005). The slope of the trend component was assessed using Sen’s method (Sen 1968), which provides a robust estimate of the linear trend and its 95% confidence range. This linear trend calculation was also performed for the time series for each spatial grid point.
A hybrid system of data-driven approaches for simulating residential energy demand profiles
Published in Journal of Building Performance Simulation, 2021
Sandhya Patidar, David Paul Jenkins, Andrew Peacock, Peter McCallum
In addition to these widely applied mathematical approaches, a range of statistical methods has been developed for de-constituting time series into a set of non-observables (latent) components (Dagum 1980). Time series are often considered as composed of four components: (i) Long-term trends; (ii) Cyclic-trends – often considered as long-term trends; (iii) Seasonal movements, and (iv) Residual/random variations (Persons 1919). These four components are assumed to be mutually independent in an additive decomposition model (represented in Equation (1)) and if there is dependence among the latent components then it can be defined through a multiplicative decomposition model (represented in Equation (2)), where is the observed time series, is the long-term trend, is the cyclic trend, is the seasonality and is the irregular/randomness. One of the widely applied approaches for extracting these latent components from time series is Classical Decomposition (CD). The CD approach utilises moving average procedure to extract trend cycles, which is then followed by extraction of various seasonal indices (a constant seasonal factor corresponding to a specified period). The CD approach has several limitations and therefore a range of subsequent version was proposed. These include X-11 seasonal adjustment (Shiskin, Young, and Musgrave 1967, February, Ladiray and Quenneville 2001), which is followed by X-11-ARIMA (Dagum 1980; Dagum 1999), X-12-ARIMA (Findley et al. 1998) and Seasonal Extraction in ARIMA Time Series (SEATS) approach (Dagum and Bianconcini 2016). The SEATS decomposition approach is mainly suitable for monthly and quarterly series only. A detailed overview of these various CD-based approaches can be referred to elsewhere (Dagum and Bianconcini 2016).