Online signal extraction by robust regression in moving windows with data-adaptive width selection (original) (raw)
Abstract
Online (also ‘real-time’ or ‘sequential’) signal extraction from noisy and outlier-interfered data streams is a basic but challenging goal. Fitting a robust Repeated Median (Siegel in Biometrika 69:242–244, 1982) regression line in a moving time window has turned out to be a promising approach (Davies et al. in J. Stat. Plan. Inference 122:65–78, 2004; Gather et al. in Comput. Stat. 21:33–51, 2006; Schettlinger et al. in Biomed. Eng. 51:49–56, 2006). The level of the regression line at the rightmost window position, which equates to the current time point in an online application, is then used for signal extraction. However, the choice of the window width has a large impact on the signal extraction, and it is impossible to predetermine an optimal fixed window width for data streams which exhibit signal changes like level shifts and sudden trend changes. We therefore propose a robust test procedure for the online detection of such signal changes. An algorithm including the test allows for online window width adaption, meaning that the window width is chosen w.r.t. the current data situation at each time point. Comparison studies show that our new procedure outperforms an existing Repeated Median filter with automatic window width selection (Schettlinger et al. in Int. J. Adapt. Control Signal Process. 24:346–362, 2010).
Access this article
Subscribe and save
- Get 10 units per month
- Download Article/Chapter or eBook
- 1 Unit = 1 Article or 1 Chapter
- Cancel anytime Subscribe now
Buy Now
Price excludes VAT (USA)
Tax calculation will be finalised during checkout.
Instant access to the full article PDF.
Similar content being viewed by others
Explore related subjects
Discover the latest articles, news and stories from top researchers in related subjects.
References
- Bernholt, T., Fried, R.: Computing the update of the repeated median regression line in linear time. Inf. Process. Lett. 88(3), 111–117 (2003)
Article MATH MathSciNet Google Scholar - Bernholt, T., Fried, R., Gather, U., Wegener, I.: Modified repeated median filters. Stat. Comput. 16, 177–192 (2006)
Article MathSciNet Google Scholar - Borowski, M., Schettlinger, K., Gather, U.: Multivariate real time signal processing by a robust adaptive regression filter. Commun. Stat., Simul. Comput. 38(2), 426–440 (2009)
Article MATH MathSciNet Google Scholar - Chen, C., Chan, J., Gerlach, R., Hsieh, W.: A comparison of estimators for regression models with change points. Stat. Comput. 21(3), 395–414 (2011)
Article MATH MathSciNet Google Scholar - Davies, P., Gather, U.: Breakdown and groups. Ann. Stat. 33(3), 977–1035 (2005)
Article MATH MathSciNet Google Scholar - Davies, P., Fried, R., Gather, U.: Robust signal extraction for on-line monitoring data. J. Stat. Plan. Inference 122, 65–78 (2004). Special Issue: Contemporary Data Analysis: Theory and Methods in Honor of John W. Tukey
Article MATH MathSciNet Google Scholar - Fried, R.: Robust filtering of time series with trends. J. Nonparametr. Stat. 16(3–4), 313–328 (2004)
Article MATH MathSciNet Google Scholar - Fried, R.: On the robust detection of edges in time series filtering. Comput. Stat. Data Anal. 52, 1063–1074 (2007)
Article MATH Google Scholar - Fried, R., Bernholt, T., Gather, U.: Repeated median and hybrid filters. Comput. Stat. Data Anal. 50(9), 2313–2338 (2006)
Article MATH MathSciNet Google Scholar - Fried, R., Einbeck, J., Gather, U.: Weighted repeated median smoothing and filtering. J. Am. Stat. Assoc. 102, 1300–1308 (2007)
Article MATH MathSciNet Google Scholar - Fried, R., Schettlinger, K., Borowski, M.: Getting started with robfilter (2012a). http://www.statistik.tu-dortmund.de/1543.html
- Fried, R., Schettlinger, K., Borowski, M., robfilter: Robust time series filters (2012b). http://CRAN.R-project.org/package=robfilter, with manual
- Gather, U., Fried, R.: Methods and algorithms for robust filtering. In: Proceedings in Computational Statistics (COMPSTAT 2004), pp. 159–170. Physica, Heidelberg (2004)
Chapter Google Scholar - Gather, U., Schettlinger, K., Fried, R.: Online signal extraction by robust linear regression. Comput. Stat. 21, 33–51 (2006)
Article MATH MathSciNet Google Scholar - Gelper, S., Schettlinger, K., Croux, C., Gather, U.: Robust online scale estimation in time series: a model-free approach. J. Stat. Plan. Inference 139(2), 335–349 (2009)
Article MATH MathSciNet Google Scholar - Rousseeuw, P., Hubert, M.: Regression-free and robust estimation of scale for bivariate data. Comput. Stat. Data Anal. 21(1), 67–85 (1996)
Article MATH MathSciNet Google Scholar - Rousseeuw, P., Leroy, A.: Robust Regression and Outlier Detection. Wiley, New York (1987)
Book MATH Google Scholar - Schettlinger, K.: Signal and variability extraction for online monitoring in intensive care. Ph.D. thesis, Faculty of Statistics, TU Dortmund University (2009)
- Schettlinger, K., Fried, R., Gather, U.: Robust filters for intensive care monitoring: beyond the running median. Biomed. Eng. 51, 49–56 (2006)
Article Google Scholar - Schettlinger, K., Fried, R., Gather, U.: Real time signal processing by adaptive repeated median filters. Int. J. Adapt. Control Signal Process. 24, 346–362 (2010)
MATH MathSciNet Google Scholar - Siegel, A.: Robust regression using repeated medians. Biometrika 69, 242–244 (1982)
Article MATH Google Scholar - Ylvisaker, D.: Test resistance. J. Am. Stat. Assoc. 72, 551–556 (1977)
Article MathSciNet Google Scholar
Acknowledgements
The financial support of the Deutsche Forschungsgemeinschaft (SFB 823, Statistical modelling of nonlinear dynamic processes) is gratefully acknowledged. The authors thank the referees for their helpful comments.
Author information
Authors and Affiliations
- Faculty of Statistics, TU Dortmund University, 44221, Dortmund, Germany
Matthias Borowski & Roland Fried
Authors
- Matthias Borowski
You can also search for this author inPubMed Google Scholar - Roland Fried
You can also search for this author inPubMed Google Scholar
Corresponding author
Correspondence toMatthias Borowski.
Appendices
Appendix A: Monte Carlo approximation of \(\hat{\mbox {V}} ( \ell,1 )\) and \(\hat{\mbox {V}} ( r,1 )\)
We approximate \(v_{n} := \hat{\mbox {V}} ( n,1 )\) for _n_=5,…,300 by the empirical variance of RM slopes which are estimated on samples coming from
Xt−n+i=mut+betatcdot(i−n)+varepsilont−n+i,quadi=1,ldots,n,X_{t-n+i} = \mu_t + \beta_t \cdot(i-n) + \varepsilon_{t-n+i},\quad i = 1,\ldots,n,Xt−n+i=mut+betatcdot(i−n)+varepsilont−n+i,quadi=1,ldots,n,
with standard normal i.i.d. errors ε t_−_n+i ∼N(0,1). W.l.o.g. we set μ t =β t =0 because of the regression equivariance of the RM slope (Rousseeuw and Leroy 1987), i.e. X t_−_n+i =ε t_−_n+i .
We generate time series (x t ) consisting of 100,000+300−1=100,299 observations. Then for each _n_=5,…,300 we move a time window {t_−_n+1…,t} over the time series, starting at time point _t_=300. Hence, for each n we obtain 100,000 RM slopes, and v n is the empirical variance computed on these 100,000 RM slopes. Due to this time series design the RM slopes are autocorrelated, as they are in practice. However, in another simulation study we also approximated the RM slope variance for independent samples. These estimates are comparable to those obtained by the time series design.
As was to be expected, the variance of the RM slope decreases monotonically with increasing window size n, see Fig. 9. In order to obtain approximations v n for _n_>300, we model the relationship between n and v n and find that the function
v(n)=4.77cdot10−7+17.71cdotn−3v(n)= 4.77 \cdot10^{-7} + 17.71 \cdot n^{-3}v(n)=4.77cdot10−7+17.71cdotn−3
gives an appropriate fit with standard error 0.0004 and coefficient of determination 0.9983.
Fig. 9
Empirical variance of 100,000 RM slopes estimated on standard normal samples of size _n_=5,…,50
Appendix B: The empirical distribution of the SCARM test statistic
This Monte Carlo study analyzes the distribution of the SCARM test statistic T t under the null hypothesis. That is, we compute T t on samples x t that come from model (4):
where we set μ t =β t =0 w.l.o.g. due to the regression equivariance of the RM slope. First of all we consider standard normal errors, as assumed in the theoretical development of the SCARM test, i.e. η t_−_n+i =0 and \(\sigma_{t} \, \varepsilon_{t-n+i} \sim N(0,\sigma_{t}^{2})\), and w.l.o.g. we set σ t =1. We generated 10,000 samples of length _n_=ℓ+r for _r_∈{5,10,…,100} and _ℓ_∈{r,r+5,…,100}. Thus, for any combination ℓ,r we obtained 10,000 realizations of the SCARM test statistic T t .
We find that the distribution of the SCARM test statistic can be well approximated by a _t_-distribution with f degrees of freedom, where f depends on ℓ and r. For each combination (ℓ,r), we compare the empirical α_- and (1−_α)-quantiles, _α_=0.01,0.02,…,0.05, of the SCARM test statistic to the corresponding theoretical quantiles of a _t_-distribution with degrees of freedom _f_=0.1,0.2,…,100, in order to find a suitable f for each combination (ℓ,r). For each (ℓ,r)-combination, we choose that f that minimizes the mean absolute difference between the empirical and the theoretical quantiles. Table 4 lists the suitable degrees of freedom f(ℓ,r) for _r_∈{5,10,…,50} and ℓ_∈{r,r+5,…,50}. The degrees of freedom f, and thus the quantiles t f,α/2 and t f,1−_α/2, are expected to be monotonically increasing in ℓ and r. However, this is not true for the approximations of f(ℓ,r) in Table 4. Therefore, for _r_∈{5,…,100} and _ℓ_∈{r,…,100} we set
f(ell,r)=minlimitsell′geqellminlimitsr′geqrbiglfbigl(ell′,r′bigr)bigr,f(\ell,r) = \min\limits _{\ell' \geq\ell} \min\limits _{r' \geq r} \bigl\{ f\bigl( \ell',r'\bigr) \bigr\},f(ell,r)=minlimitsell′geqellminlimitsr′geqrbiglfbigl(ell′,r′bigr)bigr,
(20)
with _r_′∈{5,10,…,50} and _ℓ_′∈{_r_′,r_′+5,…,50} to achieve monotonic degrees of freedom f and thus monotonic critical values t f,α/2 and t f,1−_α/2. By taking the minimum in (20) we decide for larger absolute critical values in order that the test keeps the chosen level of significance α. If ℓ or r is larger than 100, we use standard normal quantiles as critical values.
Table 4 Approximation of the SCARM test statistic distribution by a _t_-distribution: suitable degrees of freedom f(ℓ,r)
2.1 B.1 Other error types
We further investigate the distribution of T t for heavy-tailed, skewed, and contaminated errors ξ t_−_n+i , in particular:
- Noise type 1: heavy-tailed errors from a standardized _t_-distribution with three degrees of freedom;
- Noise type 2: skewed errors from a standardized Weibull distribution with scale and shape parameter two and one;
- Noise type 3: standard normal errors with 10 % contamination from N(10,1);
- Noise type 4: standard normal errors with 10 % contamination from \(N(0,\sigma_{t}^{2}=100)\).
Table 5 gives the empirical (1−α/2)-quantiles, _α_=0.01,0.005,0.001, of the computed SCARM test statistics for the four noise types and for different combinations (ℓ,r). Furthermore, the table lists the quantiles of the t f(ℓ,r)-distribution which are used as critical values for test decision. The empirical quantiles are generally lower than the t f -quantiles that are used for test decision, except for (ℓ,r)=(10,10) and given the skewed noise type 2. That is, the test keeps the chosen level of significance, even if the noise is heavy-tailed or contaminated. However, if the noise is skewed, ℓ and r should both not be too small.
Table 5 Quantiles of a t f(ℓ,r) distribution and empirical quantiles of the SCARM test statistic computed on different types of noise
Rights and permissions
About this article
Cite this article
Borowski, M., Fried, R. Online signal extraction by robust regression in moving windows with data-adaptive width selection.Stat Comput 24, 597–613 (2014). https://doi.org/10.1007/s11222-013-9391-7
- Received: 17 November 2011
- Accepted: 11 March 2013
- Published: 10 April 2013
- Issue Date: July 2014
- DOI: https://doi.org/10.1007/s11222-013-9391-7