Bootstraps

[Claire]

First, this forum has been a great resource - thank you!

I have a quick question about the bootstrapping function, in both sailfish and salmon.  Are the bootstraps being used solely to make uncertainty estimates for use in downstream tools such as sleuth, or are the abundance estimates also updated after the bootstrapping is performed?

[Rob]

Hi Claire,

  The bootstraps (or Gibbs samples) are currently only output as auxiliary information, with the goal of helping to quantify uncertainty.  However, we have looked at how summary statistics of these samples behave compared to the "main" estimates.  It does seem that the mean of the bootstraps (or posterior Gibbs samples) acts as a summary statistic that is often at least as accurate as the "main" quantification result, though differences are usually rather small.

Best,

Rob

[Claire]

Thanks for the quick reply, Rob - that was exactly the information I was looking for.