Print

Print
HI Don,

Indeed, I have tiny residuals at that voxel (0.00003). I hadn't seen that before because of the color scale where I was looking.

As for the surrounding values, I have T values that are much more reasonable - in the 7-15 range, and residuals between ~0.01 to 0.09. 

So the question now is, is there a way around this such that I can get an idea of what the data actually look like that are in the normal range? Obviously none of these are going to be final analyses. And the other thing I would like to try to make certain of is that there are no other issues other than a small sample size. Do you have any thoughts about that?

Thank you again,

~Theresa

On Oct 10, 2014, at 7:55 PM, MCLAREN, Donald <[log in to unmask]> wrote:

So, you don't have any extreme values, but all your values are very close together. This means that you have a very small residual value. When you divide the beta by a small value, you get a very big value.

What do the values around the peak look like?

Best Regards, Donald McLaren
=================
D.G. McLaren, Ph.D.
Research Fellow, Department of Neurology, Massachusetts General Hospital and
Harvard Medical School
Postdoctoral Research Fellow, GRECC, Bedford VA
Website: http://www.martinos.org/~mclaren
Office: (773) 406-2464
=====================
This e-mail contains CONFIDENTIAL INFORMATION which may contain PROTECTED
HEALTHCARE INFORMATION and may also be LEGALLY PRIVILEGED and which is
intended only for the use of the individual or entity named above. If the
reader of the e-mail is not the intended recipient or the employee or agent
responsible for delivering it to the intended recipient, you are hereby
notified that you are in possession of confidential and privileged
information. Any unauthorized use, disclosure, copying or the taking of any
action in reliance on the contents of this information is strictly
prohibited and may be unlawful. If you have received this e-mail
unintentionally, please immediately notify the sender via telephone at (773)
406-2464 or email.

On Fri, Oct 10, 2014 at 4:53 PM, Theresa M Desrochers <[log in to unmask]> wrote:
Hi Don,

Thank you for your very quick reply.

Yes, a one-sample test. 

I thought exactly the same thing, but I looked at the individual subject values for that voxel. I'm pasting in the graphic that I get from using the "plot" button (plot, fitted responses, adjusted, an explanatory variable), which should be the values for the individual subjects, correct?

All five values are between 1.412 and 1.424 in that plot. 

So it doesn't seem like to me that I have one wacky value (unless of course I am looking in the wrong place).

Thank you again,

~Theresa

<PastedGraphic-10.png>

On Oct 10, 2014, at 4:43 PM, MCLAREN, Donald <[log in to unmask]> wrote:

You can think of the T-statistic as being the con_ image divided by the ResMS image. It's a little more complicated than that, but the RPV doesn't factor into the T-statistic. With only 5 subjects, you likely have an extreme subject value.

Is this a one-sample t-test?

Best Regards, Donald McLaren
=================
D.G. McLaren, Ph.D.
Research Fellow, Department of Neurology, Massachusetts General Hospital and
Harvard Medical School
Postdoctoral Research Fellow, GRECC, Bedford VA
Website: http://www.martinos.org/~mclaren
Office: (773) 406-2464
=====================
This e-mail contains CONFIDENTIAL INFORMATION which may contain PROTECTED
HEALTHCARE INFORMATION and may also be LEGALLY PRIVILEGED and which is
intended only for the use of the individual or entity named above. If the
reader of the e-mail is not the intended recipient or the employee or agent
responsible for delivering it to the intended recipient, you are hereby
notified that you are in possession of confidential and privileged
information. Any unauthorized use, disclosure, copying or the taking of any
action in reliance on the contents of this information is strictly
prohibited and may be unlawful. If you have received this e-mail
unintentionally, please immediately notify the sender via telephone at (773)
406-2464 or email.

On Fri, Oct 10, 2014 at 4:34 PM, Theresa M Desrochers <[log in to unmask]> wrote:
Greetings,

I am trying to do a group analysis on 5 participants just to get a preliminary look at some data. It's a standard GLM with fixed effects on the subject level and a group (second level) analysis with subject as a random effect.

I know it's a low number of subjects, but I am getting an All > Baseline contrast that has *one* really high voxel (statistic 553.66), and consequently everything else looks rather ridiculous. Other contrasts are plagued by similar issues in different individual voxels, though with statistics that aren't quite so extreme.

I have looked through the All > Baseline contrasts, betas, and T-maps for all the individual subjects. I have looked at the betas and the residuals on the group level as well. The only map that looks like it has values that are clearly as outside the rest of the usual variation is the RPV.img on the group level. The same voxel has a really large value (0.29182), compared to others around it (generally < 0.01). I have looked at the residuals and RPV on the single-subject level an not been able to find that looks similarly out of the ordinary.

So my question is, does anyone know what is causing this? How is RPV calculated on the group level and how is it used in generating the contrast such that I'm having this problem? A pointer to a reference would be helpful if no one has a specific answer (but hopefully someone does!). No matter what model I run, I have the same issue at the same voxel, so it is something inherent to the data, and not the specific model I am running. 

I am hoping this question makes sense. Please let me know if I can provide any additional information.

Thank you!

~Theresa