A number of recent studies have examined trends in extreme temperature indices using a linear regression model based on ordinary least-squares. In this study, quantile regression was, for the first time, applied to ex...A number of recent studies have examined trends in extreme temperature indices using a linear regression model based on ordinary least-squares. In this study, quantile regression was, for the first time, applied to examine the trends not only in the mean but also in all parts of the distribution of several extreme temperature indices in China for the period 1960–2008. For China as a whole, the slopes in almost all the quantiles of the distribution showed a notable increase in the numbers of warm days and warm nights, and a significant decrease in the number of cool nights. These changes became much faster as the quantile increased. However, although the number of cool days exhibited a significant decrease in the mean trend estimated by classical linear regression, there was no obvious trend in the upper and lower quantiles. This finding suggests that examining the trends in different parts of the distribution of the time-series is of great importance. The spatial distribution of the trend in the 90 th quantile indicated that there was a pronounced increase in the numbers of warm days and warm nights, and a decrease in the number of cool nights for most of China, but especially in the northern and western parts of China, while there was no significant change for the number of cool days at almost all the stations.展开更多
Ordinary least squares(OLS) algorithm is widely applied in process measurement, because the sensor model used to estimate unknown parameters can be approximated through multivariate linear model. However, with few or ...Ordinary least squares(OLS) algorithm is widely applied in process measurement, because the sensor model used to estimate unknown parameters can be approximated through multivariate linear model. However, with few or noisy data or multi-collinearity, unbiased OLS leads to large variance. Biased estimators, especially ridge estimator, have been introduced to improve OLS by trading bias for variance. Ridge estimator is feasible as an estimator with smaller variance. At the same confidence level, with additive noise as the normal random variable, the less variance one estimator has, the shorter the two-sided symmetric confidence interval is. However, this finding is limited to the unbiased estimator and few studies analyze and compare the confidence levels between ridge estimator and OLS. This paper derives the matrix of ridge parameters under necessary and sufficient conditions based on which ridge estimator is superior to OLS in terms of mean squares error matrix, rather than mean squares error.Then the confidence levels between ridge estimator and OLS are compared under the condition of OLS fixed symmetric confidence interval, rather than the criteria for evaluating the validity of different unbiased estimators. We conclude that the confidence level of ridge estimator can not be directly compared with that of OLS based on the criteria available for unbiased estimators, which is verified by a simulation and a laboratory scale experiment on a single parameter measurement.展开更多
基金sponsored by the National Basic Research Program of China (973 Program, Grant No. 2012CB956203)the Knowledge Innovation Project of the Chinese Academy of Sciences (Grant No. KZCX2-EW-202)the Strategic Priority Research Program of the Chinese Academy of Sciences (Grant No. XDA05090100)
文摘A number of recent studies have examined trends in extreme temperature indices using a linear regression model based on ordinary least-squares. In this study, quantile regression was, for the first time, applied to examine the trends not only in the mean but also in all parts of the distribution of several extreme temperature indices in China for the period 1960–2008. For China as a whole, the slopes in almost all the quantiles of the distribution showed a notable increase in the numbers of warm days and warm nights, and a significant decrease in the number of cool nights. These changes became much faster as the quantile increased. However, although the number of cool days exhibited a significant decrease in the mean trend estimated by classical linear regression, there was no obvious trend in the upper and lower quantiles. This finding suggests that examining the trends in different parts of the distribution of the time-series is of great importance. The spatial distribution of the trend in the 90 th quantile indicated that there was a pronounced increase in the numbers of warm days and warm nights, and a decrease in the number of cool nights for most of China, but especially in the northern and western parts of China, while there was no significant change for the number of cool days at almost all the stations.
基金Supported by the National Natural Science Foundation of China （21006127） and the National Basic Research Program of China （2012CB720500）.
文摘Ordinary least squares(OLS) algorithm is widely applied in process measurement, because the sensor model used to estimate unknown parameters can be approximated through multivariate linear model. However, with few or noisy data or multi-collinearity, unbiased OLS leads to large variance. Biased estimators, especially ridge estimator, have been introduced to improve OLS by trading bias for variance. Ridge estimator is feasible as an estimator with smaller variance. At the same confidence level, with additive noise as the normal random variable, the less variance one estimator has, the shorter the two-sided symmetric confidence interval is. However, this finding is limited to the unbiased estimator and few studies analyze and compare the confidence levels between ridge estimator and OLS. This paper derives the matrix of ridge parameters under necessary and sufficient conditions based on which ridge estimator is superior to OLS in terms of mean squares error matrix, rather than mean squares error.Then the confidence levels between ridge estimator and OLS are compared under the condition of OLS fixed symmetric confidence interval, rather than the criteria for evaluating the validity of different unbiased estimators. We conclude that the confidence level of ridge estimator can not be directly compared with that of OLS based on the criteria available for unbiased estimators, which is verified by a simulation and a laboratory scale experiment on a single parameter measurement.
