[log in to unmask]"
type="cite">
Hi SPM experts,
I'm preparing to analyze my data and have a few specific questions on T-tests and F-tests, as well as more general questions on optimal second-level approaches. I've consulted various SPM resources in considering these questions but am still having trouble applying the theoretical explanations to my study. Any additional thoughts would be extremely helpful.
Conditions for my paradigm (a study of sexual preferences):
1. erotic clips of females
2. erotic clips of males
3. nature scenes (baseline option #1)
4. counting backwards task (baseline option #2)
I'm interested in looking at how two groups (men and women) might differ in how similarly they respond to erotic clips of males vs. erotic clips of females. The hypothesis I'm addressing is that women will show more bisexual arousal patterns (i.e. smaller differences between their responses to male stimuli and female stimuli, regardless of the direction of activation differences) than men. I'd like to see which voxels show the greatest differences between male and female erotic stimuli, in men and women, to address this hypothesis.
I've split up my questions into a few sections to hopefully make them easier to follow.
-----
FIRST LEVEL T-TEST QUESTIONS
Let's say I want to run first level T-tests looking at neural responses to female clips. Given the same order of conditions specified above [female, male, neutral, counting backwards], how would the following options differ?
a) T-test option without a baseline: 1 0 0 0 (i.e. modeling the female condition as 1)
b) T-test option with a baseline: 1 0 -1 0 (i.e. modeling the female condition as greater than the nature scenes)
Is there something that would be gained in looking at female clips without a baseline? Would a baseline add anything, or would it introduce variability that is not needed given the questions I am interested in?
In addition to whole brain analyses on the second level, I would also like to look at responses to male and female erotic stimuli within specific ROIs. I'd also like to perform correlation analyses with subjective participant data on this information. Would using baselines in the T-tests on the first level add anything in performing analyses such as this?
---
T-TEST vs. F-TESTS ON THE FIRST LEVEL
If I were to set up a first level T-test looking at responses to female stimuli [1 0 0 0], how would this be different from a first level F-test that uses the same contrast setup [also 1 0 0 0]?
I know F-tests are non-directional, and can be thought of as "containing" several T-tests. However, I am having trouble understanding the distinction in an applied sense when the tests have the same contrast specification.
More concretely, how would these differ:
1) T-test specified as [1 0 0 0] vs. F-test specified as [1 0 0 0]?
2) F-test specified as [1 1 0 0] vs. F-test specified using two lines simultaneously:
[1 -1 0 0
-1 1 0 0]
---
SECOND-LEVEL T AND F TESTS
Finally, I'd like to understand the best second-level analyses for
addressing the hypothesis of whether women show more similar responses
to male and female sexual stimuli than do men.
I was considering different analyses that I could perform, but I
wasn't sure if they were statistically valid:
1) Can I perform 2nd level T-tests on the results of first-level
F-tests (i.e., looking at both male>female and female>male contrasts
simultaneously)?
2) Is it be possible to take the absolute value of con images that
result from 1st level T-tests? If so, how would I do this?
---
Any help on any of the questions above or additional thoughts would be much, much appreciated.
If there's any more information about my paradigm that might be helpful in answering these questions, please let me know.
Many thanks,
- Victoria
