The Author Online Book Forums are Moving

The Author Online Book Forums will soon redirect to Manning's liveBook and liveVideo. All book forum content will migrate to liveBook's discussion forum and all video forum content will migrate to liveVideo. Log in to liveBook or liveVideo with your Manning credentials to join the discussion!

Thank you for your engagement in the AoF over the years! We look forward to offering you a more enhanced forum experience.

jollylinux (2) [Avatar] Offline
#1
For parametric evaluations of more than one comparison, the npmc is no longer a valid package in R. It appears to have the latest revision in 2008 and has since been removed by CRAN.

Installing the latest version from the archive, it generates an error that the following objects are masked _by_ '.GlovalEnv': class, var

It looks like it is trying to attach a variable called 'datasets' because after this error, that variable is still attached and further attempts at npmc results in more 'datasets' being attached.

Do you know of a workaround to either get the npmc to function or a replacement to do multiple comparisons like npmc?
Thanks
robert.kabacoff (170) [Avatar] Offline
#2
Re: Npmc Replacement
Thanks for letting me know about this. I will look into it.

Rob
robert.kabacoff (170) [Avatar] Offline
#3
Re: Npmc Replacement
I am really sorry that the npmc package has been archived.
Here is what I have so far...

Depending on the type of problem, you could run multiple two group comparisons using the Wilcoxon-Mann-Whitney statistic, and adjust the p-values
using the Holm, Hochberg, FDR or other such approach.

You can perform the Wilcoxon test using the wilcox.test() function. Then use the p.adjust() function to control for multiple comparisons.

See ?wilcox.test and ?p.adjust for details.

Let me know what you think.

Rob
jollylinux (2) [Avatar] Offline
#4
Re: Npmc Replacement
Thanks for looking into the npmc replacement. In my case, I only had 4 levels, so running the pairwise comparison using wilcox isn't a big issue.

Fortunately, my other experiments show using the kruskall test that there is not a difference, so I did not need to find which ones were different.
Thanks!
Wiltrud (1) [Avatar] Offline
#5
Re: Npmc Replacement
Would it be possible to install an older R version and run it on that? Lots of work, but might be worth it.
arcummin (2) [Avatar] Offline
#6
Re: Npmc Replacement
hello Robert, your book is really very good. It has made my life really easy.

I was hoping to use the npmc package, but just realized it's no longer available. I saw your last post on this was on: Posted: Dec 15, 2011 8:33 AM. Has there been any developments since? I just had a look around the r homepage and read a bit online, but I could not find an alternative.

Thanks a million.
arcummin (2) [Avatar] Offline
#7
Re: Npmc Replacement
hi again , Robert. It looks like the package "pgirmess" did the trick!!
thanks.
314616 (1) [Avatar] Offline
#8
The package can be downloaded from any internet archives. I was able to find it in a file repository writing the file name and tgz (I use MacOs). You can then install it manually, using the command:

#install.packages(filepath, repos = NULL, type="source")

My only concern is if the method to perform the post hoc comparisons is the most appropriate. After, reading some post it seems that there are many methods to correct for inflated alpha error. Hope this helps.