|
Thanks Jourdan, that makes sense. Any other input on what the resample package does would be welcome.
>>> Jourdan Gold < jgold@uoguelph.ca> 1/15/2009 12:12 PM >>> The SE gives you the SE of the bootstrap statistic so the square root of the sum of the squared differences between the mean of your bootsrap estimate and over n. It will not be exactly the same as the results obtained from a standard sample but by the assumptions of the bootstrap, it should be similar. You might want to check whether you (the bootstrap procedure you are using) is taking the deviations from the sample mean or from the average of the bootstrap resamples as there will be a difference in the results from these two measures.
I hope that this helps ----- Original Message ----- From: "Jim Daley" < jgdaley@gw.dec.state.ny.us> To: "Jourdan Gold" < jgold@uoguelph.ca> Sent: Thursday, January 15, 2009 11:53:46 AM GMT -05:00 US/Canada Eastern Subject: Re: [S] Resample variances
Thank you very much, but I'm not sure I understood your response. Lets say I'm bootstrapping the mean of some measure. The bootstrap gives me confidence limits and SE for that mean. However, I also need the variance for some other calculations. Would squaring the SE that the bootstrap provides give me the variance of the mean - as if I had used standard formulas to calculate it (which I don't want to do - I want to use the bootstrap results). Seems like that would make sense, but I'm not 100% sure about what the bootstrap method is returning and thought it would be safe to double check.
>>> Jourdan Gold < jgold@uoguelph.ca> 1/15/2009 11:44 AM >>> It should be. and you are using a standard variance formula, although in this case, on the average variation of the statistic. ----- Original Message ----- From: "Jim Daley" < jgdaley@gw.dec.state.ny.us> To: s-news@lists.biostat.wustl.edu Sent: Thursday, January 15, 2009 11:34:33 AM GMT -05:00 US/Canada Eastern Subject: [S] Resample variances
Hello, Simple question, I think. After running bootstrap, the boot object contains the standard error. If I wanted the variance associated with the parameter that I bootstrapped, could I just square the SE that the bootstrap calculates? Would that variance be analogous to the variance calculated by using a standard variance formula? Thanks, Jim -------------------------------------------------------------------- This message was distributed by s-news@lists.biostat.wustl.edu. To unsubscribe send e-mail to s-news-request@lists.biostat.wustl.edu with the BODY of the message: unsubscribe s-news |