THE BIVARIATE TREND OF LAKE MENDOTA.
WISCONSIN UNIV MADISON DEPT OF STATISTICS
Pagination or Media Count:
Freezing and thawing records of a Wisconsin Lake are analyzed for the presence of a warming trend. Records covering 111 years are analyzed using a bivariate normal linear regression model and a nonparametric or distribution-free method. For the distribution free analysis, a multivariate trend is said to exist if there is a stochastic ordering with time for at least one of the coordinates. A nonparametric test for multivariate trend is formulated and found to be consistent against a broad class of alternatives. In addition, the test has high efficiency relative to the parametric competitor for linear trend alternatives. Strong statistical evidence is presented for the alternative of trend using both methods. Author
- Hydrology, Limnology and Potamology
- Statistics and Probability