On false discovery rate thresholding for classification under sparsity

Neuvial, Pierre; Roquain, Étienne

doi:10.1214/12-aos1042

Cited by 31 publications

(29 citation statements)

References 28 publications

(101 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Neuvial and Roquain (2012) generalized the findings of Bogdan et al (2011) concerning ϕ LSU to a broader class of distributions of X 1 . In this sense, ϕ LSU adapts to the unknown degree of sparsity in the data.…”

Section: Definition 62 (Bogdan Et Al (2011))mentioning

confidence: 78%

“…As noted by Neuvial and Roquain (2012), the optimal rejection region σ c in (6.5) simplifies to a threshold for the data point x i itself if k = 1 and the likelihood ratio Θ is increasing in its argument x. Remarkable exceptions are sparse cases where class probabilities are highly unbalanced.…”

Section: Binary Classification Under Sparsitymentioning

confidence: 99%

“…(a) The risk measures R (T ) (ĥ m ) and R (I) (ĥ m ) from part (b) of Definition 6.1 are defined without a weighting by a cost parameter c. As argued by Neuvial and Roquain (2012), the results of Theorem 6.2 remain to hold true if such a weighting is considered in R (T ) (ĥ m ) and R (I) (ĥ m ). (b) The sparsity assumptions regarding ϑ 1 = ϑ 1 (m) in Theorems 6.1 and 6.2 are appropriate for signal detection problems, where a small amount of signals (corresponding to Y i = 1) is assumed within a huge amount of data points.…”

Section: Theorem 62 (Neuvial and Roquainmentioning

confidence: 99%

See 2 more Smart Citations

Simultaneous Statistical Inference

Dickhaus

2014

102

View full text Add to dashboard Cite

Section: Definition 62 (Bogdan Et Al (2011))mentioning

confidence: 78%

Section: Binary Classification Under Sparsitymentioning

confidence: 99%

Section: Theorem 62 (Neuvial and Roquainmentioning

confidence: 99%

See 1 more Smart Citation

Simultaneous Statistical Inference

Dickhaus

2014

102

View full text Add to dashboard Cite

“…A decision-theoretic and a Bayesian approach to these multiple decision problems with high-dimensional data is certainly not new as can be seen in [15, 21, 20, 4, 3, 17]. Other approaches are in [22, 23] and [18].…”

Section: Introductionmentioning

confidence: 99%

Bayes multiple decision functions

Peña

2013

Electron. J. Statist.

View full text Add to dashboard Cite

This paper deals with the problem of simultaneously making many (M) binary decisions based on one realization of a random data matrix X. M is typically large and X will usually have M rows associated with each of the M decisions to make, but for each row the data may be low dimensional. Such problems arise in many practical areas such as the biological and medical sciences, where the available dataset is from microarrays or other high-throughput technology and with the goal being to decide which among of many genes are relevant with respect to some phenotype of interest; in the engineering and reliability sciences; in astronomy; in education; and in business. A Bayesian decision-theoretic approach to this problem is implemented with the overall loss function being a cost-weighted linear combination of Type I and Type II loss functions. The class of loss functions considered allows for use of the false discovery rate (FDR), false nondiscovery rate (FNR), and missed discovery rate (MDR) in assessing the quality of decision. Through this Bayesian paradigm, the Bayes multiple decision function (BMDF) is derived and an efficient algorithm to obtain the optimal Bayes action is described. In contrast to many works in the literature where the rows of the matrix X are assumed to be stochastically independent, we allow a dependent data structure with the associations obtained through a class of frailty-induced Archimedean copulas. In particular, non-Gaussian dependent data structure, which is typical with failure-time data, can be entertained. The numerical implementation of the determination of the Bayes optimal action is facilitated through sequential Monte Carlo techniques. The theory developed could also be extended to the problem of multiple hypotheses testing, multiple classification and prediction, and high-dimensional variable selection. The proposed procedure is illustrated for the simple versus simple hypotheses setting and for the composite hypotheses setting through simulation studies. The procedure is also applied to a subset of a microarray data set from a colon cancer study.

show abstract

“…Amazon Basin(Benjamini and Hochberg, 1995;Hethcoat et al, 2019;Neuvial and Roquain, 2012). A 243 high DR and low FDR is clearly desirable, but these cannot be fixed independently in two-class 244 detection problems and both depend on the threshold value (Figure 2).…”

mentioning

confidence: 99%

Detecting tropical selective logging with SAR data requires a time series approach

Hethcoat

Carreiras

Edwards

et al. 2020

Preprint

View full text Add to dashboard Cite

Selective logging is the primary driver of forest degradation in the tropics and reduces the 26 capacity of forests to harbour biodiversity, maintain key ecosystem processes, sequester carbon, and 27 support human livelihoods. While the preceding decade has seen a tremendous improvement in the 28 ability to monitor forest disturbances from space, advances in forest monitoring have almost 29 universally relied on optical satellite data from the Landsat program, whose effectiveness is limited in 30 tropical regions with frequent cloud cover. Synthetic aperture radar (SAR) data can penetrate clouds 31 and have been utilized in forest mapping applications since the early 1990s, but no study has 32 exclusively used SAR data to map tropical selective logging. A detailed selective logging dataset from 33 three lowland tropical forest regions in the Brazilian Amazon was used to assess the effectiveness of 34 SAR data from Sentinel-1, RADARSAT-2 and PALSAR-2 for monitoring tropical selective logging. 35We built Random Forest models in an effort to classify pixel-based differences in logged and 36 unlogged areas. In addition, we used the BFAST algorithm to assess if a dense time series of Sentinel-37 1 imagery displayed recognizable shifts in pixel values after selective logging. Random Forest 38 classification with SAR data (Sentinel-1, RADARSAT-2, and ALOS-2 PALSAR-2) performed 39 poorly, having high commission and omission errors for logged observations. This suggests little to 40 no difference in pixel-based metrics between logged and unlogged areas for these sensors. In contrast, 41the Sentinel-1 time series analyses indicated that areas under higher intensity selective logging (> 20 42 m 3 ha -1 ) show a distinct spike in the number of pixels that included a breakpoint during the logging 43 season. BFAST detected breakpoints in 50% of logged pixels and exhibited a false alarm rate of 44 approximately 10% in unlogged forest. Overall our results suggest that SAR data can be used in time 45 series analyses to detect tropical selective logging at high intensity logging locations within the 46 Amazon (> 20 m 3 ha -1 ). These results have important implications for current and future abilities to 47 detect selective logging with freely available SAR data from SAOCOM 1A, the planned continuation 48 missions of Sentinel-1 (C and D), ALOS PALSAR-1 archives (expected to be opened for free access 49 in 2020), and the upcoming launch of NISAR. 50 51

show abstract

On false discovery rate thresholding for classification under sparsity

Cited by 31 publications

References 28 publications

Simultaneous Statistical Inference

Simultaneous Statistical Inference

Bayes multiple decision functions

Detecting tropical selective logging with SAR data requires a time series approach

Contact Info

Product

Resources

About