[Met_help] Re: question about mode analysis tool
John Halley Gotway
johnhg at rap.ucar.edu
Fri Dec 5 09:42:37 MST 2008
No, MET does not currently perform statistical significance testing for differences between 2 datasets.
But you're not alone. That is exactly the type of comparison the DTC needs the ability to perform as well. For example, while the 2007 CORETEST runs were verified using the NCEP Vx package and MET,
the output had to be post-processed in R to perform statistical significance tests on the differences. Ultimately, we'd like to add those analysis capabilities into MET - probably as new jobs in the
However, it won't be added for version 2.0 in February. But hopefully it'll be on our list of the following release. For the time being, you're stuck doing the statistical significance testing
yourself in R.
I've copied Eric Gilleland on this email who was involved in the 2007 CORETEST analysis and knows a lot about R.
wgallus at iastate.edu wrote:
> I hope this doesn't sound like a crazy question... I'm in the process
> of trying to write up for journal submission the stuff I did with mode
> and CRA the past 2 summers at DTC. I realized I should probably not
> just show the standard deviations I computed using mode analysis for my
> dataset, but also do statistical significance testing to see if the
> differences in my 2 datasets are significant. Does mode analysis allow
> me to do a simple t-test and spit out the p-value for the null hypothesis?
> Otherwise, I need to change how I do things and have it tell me the SDs
> of all cases so I can give that data to R to do the job. Right now, my
> code is averaging all the SDs.
More information about the Met_help