Dear Andrew,

Your tools are very much appreciated: thank you.

I would like to ask two questions:

1 Regarding command line NBS. I'm having trouble using the command line to run NBS, that is, using NBSrun. It has something to do with the S variable. It seems that the S variable should be created when running NBS, but when I run NBSrun after having ran NBS, an error prints:

Undefined function or variable "S".
Error in NBSrun (line 255)
[msg,stop]=errorcheck(UI,DIMS,S);
Error in nbs_meta (line 37)
NBSrun(UI)

So S isn't being created, how do you recommend implementing NBSrun from the commandline? I looked in the reference manual and the NBSrun code but was unable to find the solution. I imagine this may be a very simple fix... ideas?

2 Regarding p-values and multiple comparisons. I see that when using the t-test Statistical Test the test is always one-sided (from pg. 22 in the manual). Thus, it makes sense to run the contrast in both directions if one does not have a priori hypotheses regarding directionality (i.e., [-1,1] and [1,-1]). With that said, it seems that we are now running two statistical tests, so should we correct for multiple comparisons across two tests or perhaps approach the p-value as if it is a two-sided test? For example, we could simply multiply the p-values by two from the two contrasts to make two-sided tests. Do you have advice on this?

Many thanks for your time and help, best,

Matthew

Dear Andrew,

Did you happen to see the previous message in this chain? It would be great to get your thoughts.

Thank you

Hi Matt,

The S you refer to is simply a handle to the GUI. You should not have to worry about it at all if you are using the command line. You could try NBSrun(UI,[]), which means S=[], although this should not be necessary. Curious to know how this goes.

Regarding your 2nd question - yes, strictly speaking we should correct for the 2 multiple comparisons. You can do a two-sided test by selecting "F-test" and thereby avoid this problem. Your design matrix will be a bit different for an F-test - see the manual for details.

Andrew

Originally posted by :

Dear Andrew,

Thanks for your tip to use NBSrun(UI,[]), I'm no longer getting an error related to the S GUI handle. I also appreciate the suggestion to use an F-test for a two-sided test, thus avoiding multiple comparisons.

Thanks!

Glad to have signed up in this forum.  I scanned the topics and I know I will learn a lot.