| The main goal of this study is to extend the work of Woodward, Bottone, and Gray (1997) on testing for time series trend using a linear trend model. A new procedure is proposed that allows for a more flexible nonparametric specification of the form of the trend. In this new procedure, the trend is first estimated using a nonparametric regression. Bootstrap procedures are then used to estimate the null distribution of the test statistic to perform a significance test for the estimated trend. Simulations are performed to find a combination of trend estimator, test statistic, and bootstrap procedure that result in a testing procedure that has favorable size and power properties. These show that several procedures have good size and power properties that are superior to those of several other current testing procedures. The recommended testing procedure is also applied to a real dataset on global temperatures, yielding evidence of a monotonic rising trend in temperature anomalies. |
