The tenth IMSC, Beijing, China, 2007 - International Meetings on ...
The tenth IMSC, Beijing, China, 2007 - International Meetings on ...
The tenth IMSC, Beijing, China, 2007 - International Meetings on ...
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
Penalized maximal t-test for detecting undocumented mean change<br />
Speaker: Xiaolan L. Wang<br />
Xiaolan L. Wang<br />
Climate Research Divisi<strong>on</strong>, ASTD, STB, Envir<strong>on</strong>ment Canada<br />
Xiaolan.Wang@ec.gc.ca<br />
Qiuzi H. Wen and Yuehua Wu<br />
Department of Mathematics and Statistics, York University, Canada<br />
In this paper, a penalized maximal t-test (PMT) is proposed for detecting undocumented<br />
mean-shifts in climate data series. PMT takes the relative positi<strong>on</strong> of each candidate<br />
changepoint into account, to diminish the effect of unequal sample sizes <strong>on</strong> the power of<br />
detecti<strong>on</strong>. M<strong>on</strong>te Carlo simulati<strong>on</strong> studies are c<strong>on</strong>ducted to evaluate the performance of PMT,<br />
in comparis<strong>on</strong> with the most popularly used method, the standard normal homogeneity test<br />
(SNHT). An applicati<strong>on</strong> of the two methods to atmospheric pressure series recorded at a<br />
Canadian site is also presented.<br />
It is shown that the false alarm rate of PMT is very close to the specified level of<br />
significance and basically evenly distributed across all candidate changepoints, while that of<br />
SNHT can be up to 10 times higher than the specified level for points near the ends of series<br />
and much lower for the middle points. In comparis<strong>on</strong> with SNHT, c<strong>on</strong>sequently, PMT has<br />
higher power for detecting all changepoints that are not too close to the ends of series, and<br />
lower power for detecting changepoints that are near the ends of series. On average, however,<br />
PMT has significantly higher power of detecti<strong>on</strong>. <str<strong>on</strong>g>The</str<strong>on</strong>g> smaller the shift magnitude Δ relative to<br />
the noise standard deviati<strong>on</strong> σ , the greater the improvement of PMT over SNHT. <str<strong>on</strong>g>The</str<strong>on</strong>g><br />
improvement in hit rate can be as much as 14-25% for detecting small shifts (Δ < σ) regardless<br />
of time series length, and up to 5% for detecting medium shifts (Δ = σ ~ 1.5σ) in time series of<br />
length N < 100. For all detectable shift sizes, the largest improvement is always obtained when<br />
N < 100, which is of great practical importance, because most annual climate data series is of<br />
length N < 100.<br />
Penalized maximal F-test for detecting undocumented mean-shift<br />
Speaker: Xiaolan L. Wang<br />
Xiaolan L. Wang<br />
Climate Research Divisi<strong>on</strong>, ASTD, STB, Envir<strong>on</strong>ment Canada<br />
Xiaolan.Wang@ec.gc.ca<br />
In this study, a penalized maximal F-test (PMFT) is proposed for detecting undocumented<br />
mean-shifts that are not accompanied by any sudden change in the linear trend of time series.<br />
PMFT aims to even out the uneven distributi<strong>on</strong> of false alarm rate and detecti<strong>on</strong> power of the<br />
corresp<strong>on</strong>ding unpenalized maximal F-test that is based <strong>on</strong> a comm<strong>on</strong> trend two-phase<br />
27