A Double Bootstrap Method to Analyze an Intervention Time Series Model with Autoregressive Error Terms, (2000), Psychological Methods, 5, 87-101 ,S. McKnight, J.W. McKean and B. Huitema.

This procedure is intended for use on a model of the form: Y = X*b + Error, where the errors follow an autoregressive time series of order p.

The double bootstrap procedure is a two stage procedure.

• The first stage uses a bootstrap approach to obtain bias adjusted estimates of the autoregressive parameters. These estimates are used to transform the data to remove the autoregressive nature of the errors.



• The second stage uses a bootstrap method to estimate the standard errors of the parameter estimates of the model. The proposed approach is especially relevant in the behavioral sciences to analyze small sample time-series intervention, but it is appropriate for a wide class of small sample linear model problems in which there is interest in inferential statements regarding all regression parameters and autoregressive parameters in the model.

The input format for the data is X:Y matrix. The autocorrelation structure of the error term must be specified, the default is 1. This can be changed by using the select button below. You can also specifiy whether or not a bootstrap variance estimate of rho hat is desired. The default is yes, this can also be changed by selecting a different option below. If a bootstrap estimate of rho hat is selected, then CI for the AR parameters and the variance-covariance matrix of AR parameters will be calculated.

Currently the maximum number of parameters is 10 and the maximum sample size is 300. If you wish to use this software to run a larger model, please contain Joe McKean via e-mail, joe@stat.wmich.edu.
An example input data set is provided below the submit button.