 |
 |
>> 3. After naming the five contrasts to set up the F-test, what would have happened if we had selected contrasts 3 and 4 in the F-test?
>I'm not sure what you are referring to here.
On p.3 of the practical. At the end of the paragraph below the five contrasts, it reads "Next set up an F.test. Set....
Take a moment to think about what you think would have happened if we had selected contrasts 3 and 4 for our F-test instead".>In this case you would get the same result from doing an F-test with contrasts 3 and 4 as you would >get with contrasts 1 and 2. This is because the combinations of either pair can represent any two >elements vector.
Are you saying: b1[1 0] + b2[0 1] or b1[1 1] + b2[-1 1] can represent any two elements vector? I am confused. As [1 0] and [0 1] are basis vectors, a linear combination of these two vectors could generate/represent any two dimensional vectors. [1 1] and [-1 1] are not basis vectors. Perhaps I misunderstood that you wrote. Could you please explain?
>It seems that 5.0.4 does produce this conservative estimate for the highpass filter with this practical >data. So you have done nothing wrong, it is just an overly conservative output, which would not >cause a problem in practice, but is fixed in 5.0.5.>> 5. According to the practical, after I have pressed "Estimate high pass filter" and returned to the Data-tab, I was supposed to see the value changed to 90 seconds. However, FSL returned 1908. I have repeated the steps several times but still got the same inconsistency. Could you please let me know what went wrong?
>You may have an older version of FSL, as this has changed. For very short designs (like used in the practical) it could be extremely conservative. In normal length designs it should be fine.
Shall I manually change the returned value from 1908 to 90?
Thank you once again.
Date: Fri, 15 Nov 2013 00:25:59 +0000
From: [log in to unmask]
Subject: Re: [FSL] Questions about FEAT 1 Practical
To: [log in to unmask]
Hi,Say you had one EV modelling activations to viewing faces and another for activations due to viewing houses. The [1 1] contrast would test whether the average activation due to faces and houses was greater than zero or not.>> 2. On page 3, what does Mean [1 1] really mean? In the powerpoint, it states "Mean activation". What is that?
>It is the mean (average) activation across the two different conditions (each condition being associated with one EV).
Could you please give an example?
>> 3. After naming the five contrasts to set up the F-test, what would have happened if we had selected contrasts 3 and 4 in the F-test?
>I'm not sure what you are referring to here.
On p.3 of the practical. At the end of the paragraph below the five contrasts, it reads "Next set up an F.test. Set....
Take a moment to think about what you think would have happened if we had selected contrasts 3 and 4 for our F-test instead".In this case you would get the same result from doing an F-test with contrasts 3 and 4 as you would get with contrasts 1 and 2. This is because the combinations of either pair can represent any two elements vector.It seems that 5.0.4 does produce this conservative estimate for the highpass filter with this practical data. So you have done nothing wrong, it is just an overly conservative output, which would not cause a problem in practice, but is fixed in 5.0.5.>> 5. According to the practical, after I have pressed "Estimate high pass filter" and returned to the Data-tab, I was supposed to see the value changed to 90 seconds. However, FSL returned 1908. I have repeated the steps several times but still got the same inconsistency. Could you please let me know what went wrong?
>You may have an older version of FSL, as this has changed. For very short designs (like used in the practical) it could be extremely conservative. In normal length designs it should be fine.
We generally recommend that people use the latest version.My version of FSL is 5.0.4. Is it recommended to upgrade to the latest version?
All the best,Mark
I haven't following up all the posts lately. It seems that there have been some issues with the latest version.
Thank you.
> Date: Mon, 11 Nov 2013 10:59:26 +0000
> From: [log in to unmask]
> Subject: Re: [FSL] Questions about FEAT 1 Practical
> To: [log in to unmask]
>
> Hi,
>
> > Could anybody please let me know the answers? Your help is greatly appreciated. Thank you.
> >
> > 1. It seems to me that in FSL, the cutoff frequency is in period [s] rather than in Hz. Am I right?
>
> Yes.
>
> > 2. On page 3, what does Mean [1 1] really mean? In the powerpoint, it states "Mean activation". What is that?
>
> It is the mean (average) activation across the two different conditions (each condition being associated with one EV).
>
> > 3. After naming the five contrasts to set up the F-test, what would have happened if we had selected contrasts 3 and 4 in the F-test?
>
> I'm not sure what you are referring to here.
>
> > 4. In the design matrix, time goes down with every 10TRs ticked off on the left. The red bar shows the width of the highpass filter. Each TR is 4.2 sec and the red bar occupies 2.5TRs. So, the cutoff is 2.5*4.2s = 10.5s but the cutoff for high pass filter is set to 100s. I am confused. Could you please clarify?
>
> 10 TRs = 10 * 4.2 seconds = 42 seconds. Therefore 2.5 red ticks = 2.5 * 42 = 105 seconds (which would be well within your measurement error of estimating the size as 2.5 ticks).
>
> > 5. According to the practical, after I have pressed "Estimate high pass filter" and returned to the Data-tab, I was supposed to see the value changed to 90 seconds. However, FSL returned 1908. I have repeated the steps several times but still got the same inconsistency. Could you please let me know what went wrong?
>
> You may have an older version of FSL, as this has changed. For very short designs (like used in the practical) it could be extremely conservative. In normal length designs it should be fine.
>
> All the best,
> Mark