Bootstrap simulation
Bootstrap simulation (Efron and Tibshirani, 1994; Shumway and Stoffer, 2011) of the residual time series was performed
using the arima.sim procedure in R and by randomly re-sampling the innovation residuals. Using the current parameter estimates,
the covariate terms were then added to the simulated residuals via expression (1) to generate a single bootstrap sample
of the original temperature anomaly time series. Each of the monthly simulated series had to also be mean-adjusted to
ensure that, like the original series, their 20th century average was zero. To realistically represent uncertainty in the parameter
estimates a double bootstrapping procedure was performed. The first bootstrap sample was used to simulate the parameter
estimates, and the second bootstrap was used to simulate the time series using the first bootstrap parameter estimates.
The various climate statistics were then calculated, e.g. the length of the longest sequence with mean monthly temperature
exceeding the 20th century average of the corresponding month. By repeating the bootstrap simulation procedure a large
number of times the distribution of such statistics was estimated.
Bootstrap simulation
Bootstrap simulation (Efron and Tibshirani, 1994; Shumway and Stoffer, 2011) of the residual time series was performed
using the arima.sim procedure in R and by randomly re-sampling the innovation residuals. Using the current parameter estimates,
the covariate terms were then added to the simulated residuals via expression (1) to generate a single bootstrap sample
of the original temperature anomaly time series. Each of the monthly simulated series had to also be mean-adjusted to
ensure that, like the original series, their 20th century average was zero. To realistically represent uncertainty in the parameter
estimates a double bootstrapping procedure was performed. The first bootstrap sample was used to simulate the parameter
estimates, and the second bootstrap was used to simulate the time series using the first bootstrap parameter estimates.
The various climate statistics were then calculated, e.g. the length of the longest sequence with mean monthly temperature
exceeding the 20th century average of the corresponding month. By repeating the bootstrap simulation procedure a large
number of times the distribution of such statistics was estimated.
การแปล กรุณารอสักครู่..
![](//thimg.ilovetranslation.com/pic/loading_3.gif?v=b9814dd30c1d7c59_8619)