Inference - FSL
Transcript of Inference - FSL
Inference how surprising is your statistic? (thresholding)
But ... can I trust it?
Outline
• Null-hypothesis and Null-distribution
• Multiple comparisons and Family-wise error
• Different ways of being surprised
• Voxel-wise inference (Maximum z)
• Cluster-wise inference (Maximum size)
• Parametric vs non-parametric tests
• Enhanced clusters
• FDR - False Discovery Rate
Parametric vs non-parametric• As we described earlier, one of the
great things about for example the t-test is that we know the null-distribution
• But most distributions are not that simple
• And errors are not always normal-distributed
Provided that e ~ N(0,σ2)
Example: VBM-style analysis• Our data is segmented grey matter maps
• A voxel is either grey matter, or not.
Group #1(Oxford students)
Group #2(Train spotters)
Ok!
hist(e) ~ N ?L
��1
�2
⇥
��1
�2
⇥=
�0.40.6
⇥
=
Parametric vs non-parametric
• There are approximations to the Max-z and Max-size statistics
• These are valid under certain sets of assumptions
• But can be a problem when applied outside of that set of assumptions
Cluster failure: Why fMRI inferences for spatial extenthave inflated false-positive ratesAnders Eklunda,b,c,1, Thomas E. Nicholsd,e, and Hans Knutssona,c
aDivision of Medical Informatics, Department of Biomedical Engineering, Linköping University, S-581 85 Linköping, Sweden; bDivision of Statistics andMachine Learning, Department of Computer and Information Science, Linköping University, S-581 83 Linköping, Sweden; cCenter for Medical ImageScience and Visualization, Linköping University, S-581 83 Linköping, Sweden; dDepartment of Statistics, University of Warwick, Coventry CV4 7AL, UnitedKingdom; and eWMG, University of Warwick, Coventry CV4 7AL, United Kingdom
Edited by Emery N. Brown, Massachusetts General Hospital, Boston, MA, and approved May 17, 2016 (received for review February 12, 2016)
The most widely used task functional magnetic resonance imaging(fMRI) analyses use parametric statistical methods that depend on avariety of assumptions. In this work, we use real resting-state dataand a total of 3 million random task group analyses to computeempirical familywise error rates for the fMRI software packages SPM,FSL, and AFNI, as well as a nonparametric permutation method. For anominal familywise error rate of 5%, the parametric statisticalmethods are shown to be conservative for voxelwise inferenceand invalid for clusterwise inference. Our results suggest that theprincipal cause of the invalid cluster inferences is spatial autocorre-lation functions that do not follow the assumed Gaussian shape. Bycomparison, the nonparametric permutation test is found to producenominal results for voxelwise as well as clusterwise inference. Thesefindings speak to the need of validating the statistical methods beingused in the field of neuroimaging.
fMRI | statistics | false positives | cluster inference | permutation test
Since its beginning more than 20 years ago, functional magneticresonance imaging (fMRI) (1, 2) has become a popular tool
for understanding the human brain, with some 40,000 publishedpapers according to PubMed. Despite the popularity of fMRI as atool for studying brain function, the statistical methods used haverarely been validated using real data. Validations have insteadmainly been performed using simulated data (3), but it is obviouslyvery hard to simulate the complex spatiotemporal noise that arisesfrom a living human subject in an MR scanner.Through the introduction of international data-sharing initia-
tives in the neuroimaging field (4–10), it has become possible toevaluate the statistical methods using real data. Scarpazza et al.(11), for example, used freely available anatomical images from396 healthy controls (4) to investigate the validity of parametricstatistical methods for voxel-based morphometry (VBM) (12).Silver et al. (13) instead used image and genotype data from 181subjects in the Alzheimer’s Disease Neuroimaging Initiative(8, 9), to evaluate statistical methods common in imaging ge-netics. Another example of the use of open data is our previouswork (14), where a total of 1,484 resting-state fMRI datasets fromthe 1,000 Functional Connectomes Project (4) were used as nulldata for task-based, single-subject fMRI analyses with the SPMsoftware. That work found a high degree of false positives, up to70% compared with the expected 5%, likely due to a simplistictemporal autocorrelation model in SPM. It was, however, notclear whether these problems would propagate to group studies.Another unanswered question was the statistical validity of otherfMRI software packages. We address these limitations in thecurrent work with an evaluation of group inference with the threemost common fMRI software packages [SPM (15, 16), FSL (17),and AFNI (18)]. Specifically, we evaluate the packages in theirentirety, submitting the null data to the recommended suite ofpreprocessing steps integrated into each package.The main idea of this study is the same as in our previous one
(14). We analyze resting-state fMRI data with a putative taskdesign, generating results that should control the familywise error
(FWE), the chance of one or more false positives, and empiricallymeasure the FWE as the proportion of analyses that give rise toany significant results. Here, we consider both two-sample andone-sample designs. Because two groups of subjects are randomlydrawn from a large group of healthy controls, the null hypothesisof no group difference in brain activation should be true. More-over, because the resting-state fMRI data should contain noconsistent shifts in blood oxygen level-dependent (BOLD) activity,for a single group of subjects the null hypothesis of mean zeroactivation should also be true. We evaluate FWE control for bothvoxelwise inference, where significance is individually assessed ateach voxel, and clusterwise inference (19–21), where significanceis assessed on clusters formed with an arbitrary threshold.In brief, we find that all three packages have conservative
voxelwise inference and invalid clusterwise inference, for bothone- and two-sample t tests. Alarmingly, the parametric methodscan give a very high degree of false positives (up to 70%, com-pared with the nominal 5%) for clusterwise inference. By com-parison, the nonparametric permutation test (22–25) is found toproduce nominal results for both voxelwise and clusterwise in-ference for two-sample t tests, and nearly nominal results for one-sample t tests. We explore why the methods fail to appropriatelycontrol the false-positive risk.
ResultsA total of 2,880,000 random group analyses were performed tocompute the empirical false-positive rates of SPM, FSL, andAFNI; these comprise 1,000 one-sided random analyses repeatedfor 192 parameter combinations, three thresholding approaches,
Significance
Functional MRI (fMRI) is 25 years old, yet surprisingly its mostcommon statistical methods have not been validated using realdata. Here, we used resting-state fMRI data from 499 healthycontrols to conduct 3 million task group analyses. Using this nulldata with different experimental designs, we estimate the in-cidence of significant results. In theory, we should find 5% falsepositives (for a significance threshold of 5%), but insteadwe foundthat the most common software packages for fMRI analysis (SPM,FSL, AFNI) can result in false-positive rates of up to 70%. Theseresults question the validity of a number of fMRI studies and mayhave a large impact on the interpretation of weakly significantneuroimaging results.
Author contributions: A.E. and T.E.N. designed research; A.E. and T.E.N. performed re-search; A.E., T.E.N., and H.K. analyzed data; and A.E., T.E.N., and H.K. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
Freely available online through the PNAS open access option.
See Commentary on page 7699.1To whom correspondence should be addressed. Email: [email protected].
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1602413113/-/DCSupplemental.
7900–7905 | PNAS | July 12, 2016 | vol. 113 | no. 28 www.pnas.org/cgi/doi/10.1073/pnas.1602413113
Parametric vs non-parametric
• Those approximations were based on Gaussian Random Field Theory, and was an impressive body of work
• They served us fantastically well at a time when we had little choice
• But the future is non-parametric
Parametric vs non-parametric
• We can permute the data itself to create a distribution that we can use to test our statistic.
+ Makes very few assumptions about the data
+ Works for any test statistic
A simple permutation test
We have performed an experiment
And calculated a statistic, e.g. a t-value
t = 2.27If the null-hypothesis is true, there is no
difference between the groups. That means we should be able to “re-label” the individual points without changing
anything.
One re-labelling t-value after re-labelling
t = 0.67
Let’s start collecting them
Original labelling
A simple permutation test• We can permute the data itself to create a distribution
that we can use to test our statistic.
+ Makes very few assumptions about the data
+ Works for any test statistic
Second re-labelling t-value after re-labelling
t = 1.97Original labelling
And another one
A simple permutation test• We can permute the data itself to create a distribution
that we can use to test our statistic.
+ Makes very few assumptions about the data
+ Works for any test statistic
Of the 5000 re-labellings, only 90 had a t-value > 2.27 (the original labelling).
I.e. there is only a ~1.8% (90/5000) chance of obtaining a value > 2.27 if there is no
difference between the groups
C.f. p(x 2.27) = 1.79% for t185000 re-labellings. Phew!
�
Original labelling
A simple permutation test• We can permute the data itself to create a distribution
that we can use to test our statistic.
+ Makes very few assumptions about the data
+ Works for any test statistic
And we can use this for any statistic
We compared activation by painful stimuli in two groups of 5 subjects each.
This is what we gotVery intriguing activation. t8 = 4.65
But, can we trust it?
Group 1
Group 2
2nd level model
Our group difference map
max(t)=4.65
We compared activation by painful stimuli in two groups of 5 subjects each.
This is what we gotVery intriguing activation. t8 = 4.65
And we can use this for any statistic
But, can we trust it?
Permuted model
Permuted group difference map
max(t)=8.23
We compared activation by painful stimuli in two groups of 5 subjects each.
This is what we gotVery intriguing activation. t8 = 4.65
Group 1
Group 2
And we can use this for any statistic
But, can we trust it?
2nd Permutation
2nd permuted map
max(t)=5.43
We compared activation by painful stimuli in two groups of 5 subjects each.
This is what we gotVery intriguing activation. t8 = 4.65
Group 1
Group 2
And we can use this for any statistic
But, can we trust it?
3rd Permutation
3rd permuted map
max(t)=5.84
We compared activation by painful stimuli in two groups of 5 subjects each.
This is what we gotVery intriguing activation. t8 = 4.65
Group 1
Group 2
And we can use this for any statistic
But, can we trust it?
5000 permutations
3925 permutations yielded higher
max(t)-value than original labelling.We cannot reject
the null-hypothesis.
Original labelling
We compared activation by painful stimuli in two groups of 5 subjects each.
This is what we gotVery intriguing activation. t8 = 4.65
Group 1
Group 2
And we can use this for any statistic
But, can we trust it?
But beware the “exchangeability”
• When we swap the labels of two data-points we need to make sure that they are “exchangeable”
• I will start to explain “exchangeability” through a case that is not
• But first we need to learn about covariance matrices
Height and weight of a random
sample of Swedish men
• When we swap the labels of two data-points we need to make sure that they are “exchangeable”
• I will start to explain “exchangeability” through a case that is not
• But first we need to learn about covariance matrices
Covariance matrices
Mean height ≈181 cm
Mean weight ≈79.4 kgCharacterised by two means
• When we swap the labels of two data-points we need to make sure that they are “exchangeable”
• I will start to explain “exchangeability” through a case that is not
• But first we need to learn about covariance matrices
And a covariance -
matrix⌃ =
130 5252 165
�
<latexit sha1_base64="wNq9osh/sl/AGk0S0hUVqj0DAZ8=">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</latexit><latexit sha1_base64="wNq9osh/sl/AGk0S0hUVqj0DAZ8=">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</latexit><latexit sha1_base64="wNq9osh/sl/AGk0S0hUVqj0DAZ8=">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</latexit><latexit sha1_base64="wNq9osh/sl/AGk0S0hUVqj0DAZ8=">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</latexit>
Covariance matrices
• When we swap the labels of two data-points we need to make sure that they are “exchangeable”
• I will start to explain “exchangeability” through a case that is not
• But first we need to learn about covariance matrices
⌃ =
2
4130 52 4.852 165 694.8 69 156
3
5
<latexit sha1_base64="u7qX+FLcsQ0AFfGpfHi8jKaltBM=">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</latexit><latexit sha1_base64="u7qX+FLcsQ0AFfGpfHi8jKaltBM=">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</latexit><latexit sha1_base64="u7qX+FLcsQ0AFfGpfHi8jKaltBM=">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</latexit><latexit sha1_base64="u7qX+FLcsQ0AFfGpfHi8jKaltBM=">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</latexit>
Covariance matrices
• When we swap the labels of two data-points we need to make sure that they are “exchangeable”
• I will start to explain “exchangeability” through a case that is not
• But first we need to learn about covariance matrices
⌃ =<latexit sha1_base64="vUpVJd/1WzfxWh4LeTNnHhZFX0g=">AAAB8HicbVDLSgNBEOyNrxhfUY9eBoPgKeyKoBch4CXHiOYhyRJmJ7PJkHksM7NCWPIVXjwo4tXP8ebfOEn2oIkFDUVVN91dUcKZsb7/7RXW1jc2t4rbpZ3dvf2D8uFRy6hUE9okiivdibChnEnatMxy2kk0xSLitB2Nb2d++4lqw5R8sJOEhgIPJYsZwdZJj717NhQY3aB+ueJX/TnQKglyUoEcjX75qzdQJBVUWsKxMd3AT2yYYW0Z4XRa6qWGJpiM8ZB2HZVYUBNm84On6MwpAxQr7UpaNFd/T2RYGDMRkesU2I7MsjcT//O6qY2vw4zJJLVUksWiOOXIKjT7Hg2YpsTyiSOYaOZuRWSENSbWZVRyIQTLL6+S1kU18KvB3WWlVs/jKMIJnMI5BHAFNahDA5pAQMAzvMKbp70X7937WLQWvHzmGP7A+/wBoOSPpA==</latexit><latexit sha1_base64="vUpVJd/1WzfxWh4LeTNnHhZFX0g=">AAAB8HicbVDLSgNBEOyNrxhfUY9eBoPgKeyKoBch4CXHiOYhyRJmJ7PJkHksM7NCWPIVXjwo4tXP8ebfOEn2oIkFDUVVN91dUcKZsb7/7RXW1jc2t4rbpZ3dvf2D8uFRy6hUE9okiivdibChnEnatMxy2kk0xSLitB2Nb2d++4lqw5R8sJOEhgIPJYsZwdZJj717NhQY3aB+ueJX/TnQKglyUoEcjX75qzdQJBVUWsKxMd3AT2yYYW0Z4XRa6qWGJpiM8ZB2HZVYUBNm84On6MwpAxQr7UpaNFd/T2RYGDMRkesU2I7MsjcT//O6qY2vw4zJJLVUksWiOOXIKjT7Hg2YpsTyiSOYaOZuRWSENSbWZVRyIQTLL6+S1kU18KvB3WWlVs/jKMIJnMI5BHAFNahDA5pAQMAzvMKbp70X7937WLQWvHzmGP7A+/wBoOSPpA==</latexit><latexit sha1_base64="vUpVJd/1WzfxWh4LeTNnHhZFX0g=">AAAB8HicbVDLSgNBEOyNrxhfUY9eBoPgKeyKoBch4CXHiOYhyRJmJ7PJkHksM7NCWPIVXjwo4tXP8ebfOEn2oIkFDUVVN91dUcKZsb7/7RXW1jc2t4rbpZ3dvf2D8uFRy6hUE9okiivdibChnEnatMxy2kk0xSLitB2Nb2d++4lqw5R8sJOEhgIPJYsZwdZJj717NhQY3aB+ueJX/TnQKglyUoEcjX75qzdQJBVUWsKxMd3AT2yYYW0Z4XRa6qWGJpiM8ZB2HZVYUBNm84On6MwpAxQr7UpaNFd/T2RYGDMRkesU2I7MsjcT//O6qY2vw4zJJLVUksWiOOXIKjT7Hg2YpsTyiSOYaOZuRWSENSbWZVRyIQTLL6+S1kU18KvB3WWlVs/jKMIJnMI5BHAFNahDA5pAQMAzvMKbp70X7937WLQWvHzmGP7A+/wBoOSPpA==</latexit><latexit sha1_base64="vUpVJd/1WzfxWh4LeTNnHhZFX0g=">AAAB8HicbVDLSgNBEOyNrxhfUY9eBoPgKeyKoBch4CXHiOYhyRJmJ7PJkHksM7NCWPIVXjwo4tXP8ebfOEn2oIkFDUVVN91dUcKZsb7/7RXW1jc2t4rbpZ3dvf2D8uFRy6hUE9okiivdibChnEnatMxy2kk0xSLitB2Nb2d++4lqw5R8sJOEhgIPJYsZwdZJj717NhQY3aB+ueJX/TnQKglyUoEcjX75qzdQJBVUWsKxMd3AT2yYYW0Z4XRa6qWGJpiM8ZB2HZVYUBNm84On6MwpAxQr7UpaNFd/T2RYGDMRkesU2I7MsjcT//O6qY2vw4zJJLVUksWiOOXIKjT7Hg2YpsTyiSOYaOZuRWSENSbWZVRyIQTLL6+S1kU18KvB3WWlVs/jKMIJnMI5BHAFNahDA5pAQMAzvMKbp70X7937WLQWvHzmGP7A+/wBoOSPpA==</latexit>
Covariance matrices
• You may, or may not, have seen this slide in the 1st level GLM talk.
1st level fMRI data is not exchangeable
= X �
��1
�2
�
+Gaussian noise
(temporal autocorrelation)Design MatrixData from a voxel
Regressor, Explanatory Variable (EV)
Regression parameters, Effect sizes
y
=
e
+
x1 x2This time we will look more closely
at this part
e ⇠ N(0,⌃)<latexit sha1_base64="WmKqLrvjHkdWK5/0n5uVhTSlh/k=">AAACDXicbVDLSsNAFJ3UV62vqEs3g1WoICURQZcFN11JRfuAppTJdNIOnZmEmYlQQn7Ajb/ixoUibt2782+ctBG09cCFwzn3cu89fsSo0o7zZRWWlldW14rrpY3Nre0de3evpcJYYtLEIQtlx0eKMCpIU1PNSCeSBHGfkbY/vsr89j2RiobiTk8i0uNoKGhAMdJG6ttHHkd65AcJSaGnKIfXlR/FSU+9Wzrk6KRvl52qMwVcJG5OyiBHo29/eoMQx5wIjRlSqus6ke4lSGqKGUlLXqxIhPAYDUnXUIE4Ub1k+k0Kj40ygEEoTQkNp+rviQRxpSbcN53ZpWrey8T/vG6sg8teQkUUayLwbFEQM6hDmEUDB1QSrNnEEIQlNbdCPEISYW0CLJkQ3PmXF0nrrOo6VffmvFyr53EUwQE4BBXgggtQA3XQAE2AwQN4Ai/g1Xq0nq03633WWrDymX3wB9bHN4uFmzc=</latexit><latexit sha1_base64="WmKqLrvjHkdWK5/0n5uVhTSlh/k=">AAACDXicbVDLSsNAFJ3UV62vqEs3g1WoICURQZcFN11JRfuAppTJdNIOnZmEmYlQQn7Ajb/ixoUibt2782+ctBG09cCFwzn3cu89fsSo0o7zZRWWlldW14rrpY3Nre0de3evpcJYYtLEIQtlx0eKMCpIU1PNSCeSBHGfkbY/vsr89j2RiobiTk8i0uNoKGhAMdJG6ttHHkd65AcJSaGnKIfXlR/FSU+9Wzrk6KRvl52qMwVcJG5OyiBHo29/eoMQx5wIjRlSqus6ke4lSGqKGUlLXqxIhPAYDUnXUIE4Ub1k+k0Kj40ygEEoTQkNp+rviQRxpSbcN53ZpWrey8T/vG6sg8teQkUUayLwbFEQM6hDmEUDB1QSrNnEEIQlNbdCPEISYW0CLJkQ3PmXF0nrrOo6VffmvFyr53EUwQE4BBXgggtQA3XQAE2AwQN4Ai/g1Xq0nq03633WWrDymX3wB9bHN4uFmzc=</latexit><latexit sha1_base64="WmKqLrvjHkdWK5/0n5uVhTSlh/k=">AAACDXicbVDLSsNAFJ3UV62vqEs3g1WoICURQZcFN11JRfuAppTJdNIOnZmEmYlQQn7Ajb/ixoUibt2782+ctBG09cCFwzn3cu89fsSo0o7zZRWWlldW14rrpY3Nre0de3evpcJYYtLEIQtlx0eKMCpIU1PNSCeSBHGfkbY/vsr89j2RiobiTk8i0uNoKGhAMdJG6ttHHkd65AcJSaGnKIfXlR/FSU+9Wzrk6KRvl52qMwVcJG5OyiBHo29/eoMQx5wIjRlSqus6ke4lSGqKGUlLXqxIhPAYDUnXUIE4Ub1k+k0Kj40ygEEoTQkNp+rviQRxpSbcN53ZpWrey8T/vG6sg8teQkUUayLwbFEQM6hDmEUDB1QSrNnEEIQlNbdCPEISYW0CLJkQ3PmXF0nrrOo6VffmvFyr53EUwQE4BBXgggtQA3XQAE2AwQN4Ai/g1Xq0nq03633WWrDymX3wB9bHN4uFmzc=</latexit><latexit sha1_base64="WmKqLrvjHkdWK5/0n5uVhTSlh/k=">AAACDXicbVDLSsNAFJ3UV62vqEs3g1WoICURQZcFN11JRfuAppTJdNIOnZmEmYlQQn7Ajb/ixoUibt2782+ctBG09cCFwzn3cu89fsSo0o7zZRWWlldW14rrpY3Nre0de3evpcJYYtLEIQtlx0eKMCpIU1PNSCeSBHGfkbY/vsr89j2RiobiTk8i0uNoKGhAMdJG6ttHHkd65AcJSaGnKIfXlR/FSU+9Wzrk6KRvl52qMwVcJG5OyiBHo29/eoMQx5wIjRlSqus6ke4lSGqKGUlLXqxIhPAYDUnXUIE4Ub1k+k0Kj40ygEEoTQkNp+rviQRxpSbcN53ZpWrey8T/vG6sg8teQkUUayLwbFEQM6hDmEUDB1QSrNnEEIQlNbdCPEISYW0CLJkQ3PmXF0nrrOo6VffmvFyr53EUwQE4BBXgggtQA3XQAE2AwQN4Ai/g1Xq0nq03633WWrDymX3wB9bHN4uFmzc=</latexit>
Our old friend “the covariance matrix”
• One important component of noise in fMRI consists of physiological/neuronal events convolved by the HRF
1st level fMRI data is not exchangeable
• One important component of noise in fMRI consists of physiological/neuronal events convolved by the HRF
1st level fMRI data is not exchangeable
…
…
If we sample this every 20 seconds it no longer looks “smooth”
• One important component of noise in fMRI consists of physiological/neuronal events convolved by the HRF
1st level fMRI data is not exchangeable
Variance at point 1
Variance at point 2
Covariance between points
1 and 2
e ⇠ N(0,�2I)<latexit sha1_base64="Hssh0UyRWYZvKSNZha/T9pQ4RBk=">AAACGnicbZDLSsNAFIYn9VbrLerSzWARKkhJiqDLohvdSAV7gSaWyXTSDp1JwsxEKCHP4cZXceNCEXfixrdx0qagrT8M/HznHOac34sYlcqyvo3C0vLK6lpxvbSxubW9Y+7utWQYC0yaOGSh6HhIEkYD0lRUMdKJBEHcY6TtjS6zevuBCEnD4E6NI+JyNAioTzFSGvVM2+FIDT0/ISl0JOXwpjIjVnqiyYCj+xqcsev0uGeWrao1EVw0dm7KIFejZ346/RDHnAQKMyRl17Yi5SZIKIoZSUtOLEmE8AgNSFfbAHEi3WRyWgqPNOlDPxT6BQpO6O+JBHEpx9zTndmKcr6Wwf9q3Vj5525CgyhWJMDTj/yYQRXCLCfYp4JgxcbaICyo3hXiIRIIK51mSYdgz5+8aFq1qm1V7dvTcv0ij6MIDsAhqAAbnIE6uAIN0AQYPIJn8ArejCfjxXg3PqatBSOf2Qd/ZHz9AJrcoJI=</latexit><latexit sha1_base64="Hssh0UyRWYZvKSNZha/T9pQ4RBk=">AAACGnicbZDLSsNAFIYn9VbrLerSzWARKkhJiqDLohvdSAV7gSaWyXTSDp1JwsxEKCHP4cZXceNCEXfixrdx0qagrT8M/HznHOac34sYlcqyvo3C0vLK6lpxvbSxubW9Y+7utWQYC0yaOGSh6HhIEkYD0lRUMdKJBEHcY6TtjS6zevuBCEnD4E6NI+JyNAioTzFSGvVM2+FIDT0/ISl0JOXwpjIjVnqiyYCj+xqcsev0uGeWrao1EVw0dm7KIFejZ346/RDHnAQKMyRl17Yi5SZIKIoZSUtOLEmE8AgNSFfbAHEi3WRyWgqPNOlDPxT6BQpO6O+JBHEpx9zTndmKcr6Wwf9q3Vj5525CgyhWJMDTj/yYQRXCLCfYp4JgxcbaICyo3hXiIRIIK51mSYdgz5+8aFq1qm1V7dvTcv0ij6MIDsAhqAAbnIE6uAIN0AQYPIJn8ArejCfjxXg3PqatBSOf2Qd/ZHz9AJrcoJI=</latexit><latexit sha1_base64="Hssh0UyRWYZvKSNZha/T9pQ4RBk=">AAACGnicbZDLSsNAFIYn9VbrLerSzWARKkhJiqDLohvdSAV7gSaWyXTSDp1JwsxEKCHP4cZXceNCEXfixrdx0qagrT8M/HznHOac34sYlcqyvo3C0vLK6lpxvbSxubW9Y+7utWQYC0yaOGSh6HhIEkYD0lRUMdKJBEHcY6TtjS6zevuBCEnD4E6NI+JyNAioTzFSGvVM2+FIDT0/ISl0JOXwpjIjVnqiyYCj+xqcsev0uGeWrao1EVw0dm7KIFejZ346/RDHnAQKMyRl17Yi5SZIKIoZSUtOLEmE8AgNSFfbAHEi3WRyWgqPNOlDPxT6BQpO6O+JBHEpx9zTndmKcr6Wwf9q3Vj5525CgyhWJMDTj/yYQRXCLCfYp4JgxcbaICyo3hXiIRIIK51mSYdgz5+8aFq1qm1V7dvTcv0ij6MIDsAhqAAbnIE6uAIN0AQYPIJn8ArejCfjxXg3PqatBSOf2Qd/ZHz9AJrcoJI=</latexit><latexit sha1_base64="Hssh0UyRWYZvKSNZha/T9pQ4RBk=">AAACGnicbZDLSsNAFIYn9VbrLerSzWARKkhJiqDLohvdSAV7gSaWyXTSDp1JwsxEKCHP4cZXceNCEXfixrdx0qagrT8M/HznHOac34sYlcqyvo3C0vLK6lpxvbSxubW9Y+7utWQYC0yaOGSh6HhIEkYD0lRUMdKJBEHcY6TtjS6zevuBCEnD4E6NI+JyNAioTzFSGvVM2+FIDT0/ISl0JOXwpjIjVnqiyYCj+xqcsev0uGeWrao1EVw0dm7KIFejZ346/RDHnAQKMyRl17Yi5SZIKIoZSUtOLEmE8AgNSFfbAHEi3WRyWgqPNOlDPxT6BQpO6O+JBHEpx9zTndmKcr6Wwf9q3Vj5525CgyhWJMDTj/yYQRXCLCfYp4JgxcbaICyo3hXiIRIIK51mSYdgz5+8aFq1qm1V7dvTcv0ij6MIDsAhqAAbnIE6uAIN0AQYPIJn8ArejCfjxXg3PqatBSOf2Qd/ZHz9AJrcoJI=</latexit>
• One important component of noise in fMRI consists of physiological/neuronal events convolved by the HRF
1st level fMRI data is not exchangeable
…
But that is not a realistic TR. What about every 3 seconds?
• One important component of noise in fMRI consists of physiological/neuronal events convolved by the HRF
1st level fMRI data is not exchangeable
Variance at point 1
Variance at point 2
Covariance between points
1 and 2
• Let us now return to our model again
1st level fMRI data is not exchangeable
= X �
��1
�2
�
+Gaussian noise
(temporal autocorrelation)Design MatrixData from
a voxel
Regressor, Explanatory Variable (EV)
Regression parameters, Effect sizes
y
=
e
+
x1 x2
• The model consists of our regressors X and the noise model
• All permutations must result in “equivalent models”
• Let us now see what happens if we swap two data-points (points 5 and 10)
e ⇠ N(0,⌃)<latexit sha1_base64="WmKqLrvjHkdWK5/0n5uVhTSlh/k=">AAACDXicbVDLSsNAFJ3UV62vqEs3g1WoICURQZcFN11JRfuAppTJdNIOnZmEmYlQQn7Ajb/ixoUibt2782+ctBG09cCFwzn3cu89fsSo0o7zZRWWlldW14rrpY3Nre0de3evpcJYYtLEIQtlx0eKMCpIU1PNSCeSBHGfkbY/vsr89j2RiobiTk8i0uNoKGhAMdJG6ttHHkd65AcJSaGnKIfXlR/FSU+9Wzrk6KRvl52qMwVcJG5OyiBHo29/eoMQx5wIjRlSqus6ke4lSGqKGUlLXqxIhPAYDUnXUIE4Ub1k+k0Kj40ygEEoTQkNp+rviQRxpSbcN53ZpWrey8T/vG6sg8teQkUUayLwbFEQM6hDmEUDB1QSrNnEEIQlNbdCPEISYW0CLJkQ3PmXF0nrrOo6VffmvFyr53EUwQE4BBXgggtQA3XQAE2AwQN4Ai/g1Xq0nq03633WWrDymX3wB9bHN4uFmzc=</latexit><latexit sha1_base64="WmKqLrvjHkdWK5/0n5uVhTSlh/k=">AAACDXicbVDLSsNAFJ3UV62vqEs3g1WoICURQZcFN11JRfuAppTJdNIOnZmEmYlQQn7Ajb/ixoUibt2782+ctBG09cCFwzn3cu89fsSo0o7zZRWWlldW14rrpY3Nre0de3evpcJYYtLEIQtlx0eKMCpIU1PNSCeSBHGfkbY/vsr89j2RiobiTk8i0uNoKGhAMdJG6ttHHkd65AcJSaGnKIfXlR/FSU+9Wzrk6KRvl52qMwVcJG5OyiBHo29/eoMQx5wIjRlSqus6ke4lSGqKGUlLXqxIhPAYDUnXUIE4Ub1k+k0Kj40ygEEoTQkNp+rviQRxpSbcN53ZpWrey8T/vG6sg8teQkUUayLwbFEQM6hDmEUDB1QSrNnEEIQlNbdCPEISYW0CLJkQ3PmXF0nrrOo6VffmvFyr53EUwQE4BBXgggtQA3XQAE2AwQN4Ai/g1Xq0nq03633WWrDymX3wB9bHN4uFmzc=</latexit><latexit sha1_base64="WmKqLrvjHkdWK5/0n5uVhTSlh/k=">AAACDXicbVDLSsNAFJ3UV62vqEs3g1WoICURQZcFN11JRfuAppTJdNIOnZmEmYlQQn7Ajb/ixoUibt2782+ctBG09cCFwzn3cu89fsSo0o7zZRWWlldW14rrpY3Nre0de3evpcJYYtLEIQtlx0eKMCpIU1PNSCeSBHGfkbY/vsr89j2RiobiTk8i0uNoKGhAMdJG6ttHHkd65AcJSaGnKIfXlR/FSU+9Wzrk6KRvl52qMwVcJG5OyiBHo29/eoMQx5wIjRlSqus6ke4lSGqKGUlLXqxIhPAYDUnXUIE4Ub1k+k0Kj40ygEEoTQkNp+rviQRxpSbcN53ZpWrey8T/vG6sg8teQkUUayLwbFEQM6hDmEUDB1QSrNnEEIQlNbdCPEISYW0CLJkQ3PmXF0nrrOo6VffmvFyr53EUwQE4BBXgggtQA3XQAE2AwQN4Ai/g1Xq0nq03633WWrDymX3wB9bHN4uFmzc=</latexit><latexit sha1_base64="WmKqLrvjHkdWK5/0n5uVhTSlh/k=">AAACDXicbVDLSsNAFJ3UV62vqEs3g1WoICURQZcFN11JRfuAppTJdNIOnZmEmYlQQn7Ajb/ixoUibt2782+ctBG09cCFwzn3cu89fsSo0o7zZRWWlldW14rrpY3Nre0de3evpcJYYtLEIQtlx0eKMCpIU1PNSCeSBHGfkbY/vsr89j2RiobiTk8i0uNoKGhAMdJG6ttHHkd65AcJSaGnKIfXlR/FSU+9Wzrk6KRvl52qMwVcJG5OyiBHo29/eoMQx5wIjRlSqus6ke4lSGqKGUlLXqxIhPAYDUnXUIE4Ub1k+k0Kj40ygEEoTQkNp+rviQRxpSbcN53ZpWrey8T/vG6sg8teQkUUayLwbFEQM6hDmEUDB1QSrNnEEIQlNbdCPEISYW0CLJkQ3PmXF0nrrOo6VffmvFyr53EUwQE4BBXgggtQA3XQAE2AwQN4Ai/g1Xq0nq03633WWrDymX3wB9bHN4uFmzc=</latexit>
• One important component of noise in fMRI consists of physiological/neuronal events convolved by the HRF
1st level fMRI data is not exchangeable
“Point” 10 now covaries with points 4 and 6
“Point 5” now covaries with
points 9 and 11
And the models are no longer
equivalent
• One important component of noise in fMRI consists of physiological/neuronal events convolved by the HRF
1st level fMRI data is not exchangeable
And the models are no longer
equivalent
And for a random permutation …
• Data-points are not “exchangeable” if swapping them means that the noise covariance-matrix ends up looking different.
• Formally we say that “The joint distribution of the data must be unchanged by the permutations under the null-hypothesis”.
• If the noise covariance-matrix has non-zero off-diagonal elements (covariances) you need to beware.
• You typically never estimate or see the covariance-matrix. You need to “imagine it” and determine from that if there is a problem.
Back to exchangeability
Examples of exchangeability: Two groups unpaired
This is the “exchangeability group”. Here all scans are in the same group, which means any scan can be
exchanged for any other.
N.B. The “group” labelling is used for completely
different purposes when using FLAME/GRFT
The implicit assumption here is that data from all subjects have the same uncertainty and are all
independent
Assumed covariance matrix
Examples of exchangeability: Two groups unpaired
123456789
10
Original
63785124910
Perm 1
61749583102
Perm 2 …
Examples of exchangeability: Two groups unpaired
123456789
10
Original
63785124910
Perm 1
61749583102
Perm 2
17694285310
Perm 3
N.B. Equivalent{ <latexit sha1_base64="RG03HEwJX1WBjc+2nWlsTV1XmgQ=">AAAB6XicbVBNS8NAEJ3Ur1q/oh69LBbBU0lE0GPRi8cq9gPaUDbbSbt0swm7G6GE/gMvHhTx6j/y5r9x2+agrQ8GHu/NMDMvTAXXxvO+ndLa+sbmVnm7srO7t3/gHh61dJIphk2WiER1QqpRcIlNw43ATqqQxqHAdji+nfntJ1SaJ/LRTFIMYjqUPOKMGis99PK+W/Vq3hxklfgFqUKBRt/96g0SlsUoDRNU667vpSbIqTKcCZxWepnGlLIxHWLXUklj1EE+v3RKzqwyIFGibElD5urviZzGWk/i0HbG1Iz0sjcT//O6mYmug5zLNDMo2WJRlAliEjJ7mwy4QmbExBLKFLe3EjaiijJjw6nYEPzll1dJ66LmezX//rJavyniKMMJnMI5+HAFdbiDBjSBQQTP8Apvzth5cd6dj0VrySlmjuEPnM8fm4SNZQ==</latexit><latexit sha1_base64="RG03HEwJX1WBjc+2nWlsTV1XmgQ=">AAAB6XicbVBNS8NAEJ3Ur1q/oh69LBbBU0lE0GPRi8cq9gPaUDbbSbt0swm7G6GE/gMvHhTx6j/y5r9x2+agrQ8GHu/NMDMvTAXXxvO+ndLa+sbmVnm7srO7t3/gHh61dJIphk2WiER1QqpRcIlNw43ATqqQxqHAdji+nfntJ1SaJ/LRTFIMYjqUPOKMGis99PK+W/Vq3hxklfgFqUKBRt/96g0SlsUoDRNU667vpSbIqTKcCZxWepnGlLIxHWLXUklj1EE+v3RKzqwyIFGibElD5urviZzGWk/i0HbG1Iz0sjcT//O6mYmug5zLNDMo2WJRlAliEjJ7mwy4QmbExBLKFLe3EjaiijJjw6nYEPzll1dJ66LmezX//rJavyniKMMJnMI5+HAFdbiDBjSBQQTP8Apvzth5cd6dj0VrySlmjuEPnM8fm4SNZQ==</latexit><latexit sha1_base64="RG03HEwJX1WBjc+2nWlsTV1XmgQ=">AAAB6XicbVBNS8NAEJ3Ur1q/oh69LBbBU0lE0GPRi8cq9gPaUDbbSbt0swm7G6GE/gMvHhTx6j/y5r9x2+agrQ8GHu/NMDMvTAXXxvO+ndLa+sbmVnm7srO7t3/gHh61dJIphk2WiER1QqpRcIlNw43ATqqQxqHAdji+nfntJ1SaJ/LRTFIMYjqUPOKMGis99PK+W/Vq3hxklfgFqUKBRt/96g0SlsUoDRNU667vpSbIqTKcCZxWepnGlLIxHWLXUklj1EE+v3RKzqwyIFGibElD5urviZzGWk/i0HbG1Iz0sjcT//O6mYmug5zLNDMo2WJRlAliEjJ7mwy4QmbExBLKFLe3EjaiijJjw6nYEPzll1dJ66LmezX//rJavyniKMMJnMI5+HAFdbiDBjSBQQTP8Apvzth5cd6dj0VrySlmjuEPnM8fm4SNZQ==</latexit><latexit sha1_base64="RG03HEwJX1WBjc+2nWlsTV1XmgQ=">AAAB6XicbVBNS8NAEJ3Ur1q/oh69LBbBU0lE0GPRi8cq9gPaUDbbSbt0swm7G6GE/gMvHhTx6j/y5r9x2+agrQ8GHu/NMDMvTAXXxvO+ndLa+sbmVnm7srO7t3/gHh61dJIphk2WiER1QqpRcIlNw43ATqqQxqHAdji+nfntJ1SaJ/LRTFIMYjqUPOKMGis99PK+W/Vq3hxklfgFqUKBRt/96g0SlsUoDRNU667vpSbIqTKcCZxWepnGlLIxHWLXUklj1EE+v3RKzqwyIFGibElD5urviZzGWk/i0HbG1Iz0sjcT//O6mYmug5zLNDMo2WJRlAliEjJ7mwy4QmbExBLKFLe3EjaiijJjw6nYEPzll1dJ66LmezX//rJavyniKMMJnMI5+HAFdbiDBjSBQQTP8Apvzth5cd6dj0VrySlmjuEPnM8fm4SNZQ==</latexit>
Examples of exchangeability: Two groups unpaired
Examples of exchangeability:Single group average
Here we model a single mean and want to know if that is different from zero
But there isn’t really anything to permute, or
is there?
Examples of exchangeability:Single group average
++++++++++
Original
t = �0.17<latexit sha1_base64="z0uMqgB7mpi8OruR79tohv3kRHo=">AAAB8HicbVBNS8NAEJ3Ur1q/qh69LBbBiyURoV6EgpceK9haaUPZbDft0t0k7E6EEvorvHhQxKs/x5v/xm2bg7Y+GHi8N8PMvCCRwqDrfjuFtfWNza3idmlnd2//oHx41DZxqhlvsVjGuhNQw6WIeAsFSt5JNKcqkPwhGN/O/Icnro2Io3ucJNxXdBiJUDCKVnpEckMu3KpX65crbtWdg6wSLycVyNHsl796g5ilikfIJDWm67kJ+hnVKJjk01IvNTyhbEyHvGtpRBU3fjY/eErOrDIgYaxtRUjm6u+JjCpjJiqwnYriyCx7M/E/r5tieO1nIkpS5BFbLApTSTAms+/JQGjOUE4soUwLeythI6opQ5tRyYbgLb+8StqXVc9GdndVqTfyOIpwAqdwDh7UoA4NaEILGCh4hld4c7Tz4rw7H4vWgpPPHMMfOJ8/P6qOvg==</latexit><latexit sha1_base64="z0uMqgB7mpi8OruR79tohv3kRHo=">AAAB8HicbVBNS8NAEJ3Ur1q/qh69LBbBiyURoV6EgpceK9haaUPZbDft0t0k7E6EEvorvHhQxKs/x5v/xm2bg7Y+GHi8N8PMvCCRwqDrfjuFtfWNza3idmlnd2//oHx41DZxqhlvsVjGuhNQw6WIeAsFSt5JNKcqkPwhGN/O/Icnro2Io3ucJNxXdBiJUDCKVnpEckMu3KpX65crbtWdg6wSLycVyNHsl796g5ilikfIJDWm67kJ+hnVKJjk01IvNTyhbEyHvGtpRBU3fjY/eErOrDIgYaxtRUjm6u+JjCpjJiqwnYriyCx7M/E/r5tieO1nIkpS5BFbLApTSTAms+/JQGjOUE4soUwLeythI6opQ5tRyYbgLb+8StqXVc9GdndVqTfyOIpwAqdwDh7UoA4NaEILGCh4hld4c7Tz4rw7H4vWgpPPHMMfOJ8/P6qOvg==</latexit><latexit sha1_base64="z0uMqgB7mpi8OruR79tohv3kRHo=">AAAB8HicbVBNS8NAEJ3Ur1q/qh69LBbBiyURoV6EgpceK9haaUPZbDft0t0k7E6EEvorvHhQxKs/x5v/xm2bg7Y+GHi8N8PMvCCRwqDrfjuFtfWNza3idmlnd2//oHx41DZxqhlvsVjGuhNQw6WIeAsFSt5JNKcqkPwhGN/O/Icnro2Io3ucJNxXdBiJUDCKVnpEckMu3KpX65crbtWdg6wSLycVyNHsl796g5ilikfIJDWm67kJ+hnVKJjk01IvNTyhbEyHvGtpRBU3fjY/eErOrDIgYaxtRUjm6u+JjCpjJiqwnYriyCx7M/E/r5tieO1nIkpS5BFbLApTSTAms+/JQGjOUE4soUwLeythI6opQ5tRyYbgLb+8StqXVc9GdndVqTfyOIpwAqdwDh7UoA4NaEILGCh4hld4c7Tz4rw7H4vWgpPPHMMfOJ8/P6qOvg==</latexit><latexit sha1_base64="z0uMqgB7mpi8OruR79tohv3kRHo=">AAAB8HicbVBNS8NAEJ3Ur1q/qh69LBbBiyURoV6EgpceK9haaUPZbDft0t0k7E6EEvorvHhQxKs/x5v/xm2bg7Y+GHi8N8PMvCCRwqDrfjuFtfWNza3idmlnd2//oHx41DZxqhlvsVjGuhNQw6WIeAsFSt5JNKcqkPwhGN/O/Icnro2Io3ucJNxXdBiJUDCKVnpEckMu3KpX65crbtWdg6wSLycVyNHsl796g5ilikfIJDWm67kJ+hnVKJjk01IvNTyhbEyHvGtpRBU3fjY/eErOrDIgYaxtRUjm6u+JjCpjJiqwnYriyCx7M/E/r5tieO1nIkpS5BFbLApTSTAms+/JQGjOUE4soUwLeythI6opQ5tRyYbgLb+8StqXVc9GdndVqTfyOIpwAqdwDh7UoA4NaEILGCh4hld4c7Tz4rw7H4vWgpPPHMMfOJ8/P6qOvg==</latexit>
Examples of exchangeability:Single group average
+-+--++-++
First flip
t = �0.09<latexit sha1_base64="fL5QFtG/p4x7ilO8c6QMjSKxNtk=">AAAB8HicbVBNSwMxEJ31s9avqkcvwSJ4seyKoB6EgpceK9gPaZeSTbNtaDa7JLNCWforvHhQxKs/x5v/xrTdg7Y+CHm8N8PMvCCRwqDrfjsrq2vrG5uFreL2zu7efungsGniVDPeYLGMdTughkuheAMFSt5ONKdRIHkrGN1N/dYT10bE6gHHCfcjOlAiFIyilR6R3JJzt+Le9Epl+81AlomXkzLkqPdKX91+zNKIK2SSGtPx3AT9jGoUTPJJsZsanlA2ogPesVTRiBs/my08IadW6ZMw1vYpJDP1d0dGI2PGUWArI4pDs+hNxf+8TorhtZ8JlaTIFZsPClNJMCbT60lfaM5Qji2hTAu7K2FDqilDm1HRhuAtnrxMmhcVz61495flai2PowDHcAJn4MEVVKEGdWgAgwie4RXeHO28OO/Ox7x0xcl7juAPnM8fQS2Ovw==</latexit><latexit sha1_base64="fL5QFtG/p4x7ilO8c6QMjSKxNtk=">AAAB8HicbVBNSwMxEJ31s9avqkcvwSJ4seyKoB6EgpceK9gPaZeSTbNtaDa7JLNCWforvHhQxKs/x5v/xrTdg7Y+CHm8N8PMvCCRwqDrfjsrq2vrG5uFreL2zu7efungsGniVDPeYLGMdTughkuheAMFSt5ONKdRIHkrGN1N/dYT10bE6gHHCfcjOlAiFIyilR6R3JJzt+Le9Epl+81AlomXkzLkqPdKX91+zNKIK2SSGtPx3AT9jGoUTPJJsZsanlA2ogPesVTRiBs/my08IadW6ZMw1vYpJDP1d0dGI2PGUWArI4pDs+hNxf+8TorhtZ8JlaTIFZsPClNJMCbT60lfaM5Qji2hTAu7K2FDqilDm1HRhuAtnrxMmhcVz61495flai2PowDHcAJn4MEVVKEGdWgAgwie4RXeHO28OO/Ox7x0xcl7juAPnM8fQS2Ovw==</latexit><latexit sha1_base64="fL5QFtG/p4x7ilO8c6QMjSKxNtk=">AAAB8HicbVBNSwMxEJ31s9avqkcvwSJ4seyKoB6EgpceK9gPaZeSTbNtaDa7JLNCWforvHhQxKs/x5v/xrTdg7Y+CHm8N8PMvCCRwqDrfjsrq2vrG5uFreL2zu7efungsGniVDPeYLGMdTughkuheAMFSt5ONKdRIHkrGN1N/dYT10bE6gHHCfcjOlAiFIyilR6R3JJzt+Le9Epl+81AlomXkzLkqPdKX91+zNKIK2SSGtPx3AT9jGoUTPJJsZsanlA2ogPesVTRiBs/my08IadW6ZMw1vYpJDP1d0dGI2PGUWArI4pDs+hNxf+8TorhtZ8JlaTIFZsPClNJMCbT60lfaM5Qji2hTAu7K2FDqilDm1HRhuAtnrxMmhcVz61495flai2PowDHcAJn4MEVVKEGdWgAgwie4RXeHO28OO/Ox7x0xcl7juAPnM8fQS2Ovw==</latexit><latexit sha1_base64="fL5QFtG/p4x7ilO8c6QMjSKxNtk=">AAAB8HicbVBNSwMxEJ31s9avqkcvwSJ4seyKoB6EgpceK9gPaZeSTbNtaDa7JLNCWforvHhQxKs/x5v/xrTdg7Y+CHm8N8PMvCCRwqDrfjsrq2vrG5uFreL2zu7efungsGniVDPeYLGMdTughkuheAMFSt5ONKdRIHkrGN1N/dYT10bE6gHHCfcjOlAiFIyilR6R3JJzt+Le9Epl+81AlomXkzLkqPdKX91+zNKIK2SSGtPx3AT9jGoUTPJJsZsanlA2ogPesVTRiBs/my08IadW6ZMw1vYpJDP1d0dGI2PGUWArI4pDs+hNxf+8TorhtZ8JlaTIFZsPClNJMCbT60lfaM5Qji2hTAu7K2FDqilDm1HRhuAtnrxMmhcVz61495flai2PowDHcAJn4MEVVKEGdWgAgwie4RXeHO28OO/Ox7x0xcl7juAPnM8fQS2Ovw==</latexit>
t = 1.54<latexit sha1_base64="6q819vkUgiYHlbL7dyf4a4CC5DQ=">AAAB73icbVBNS8NAEJ3Ur1q/qh69LBbBU0hE0YtQ8NJjBfsBbSib7aZdutnE3YlQSv+EFw+KePXvePPfuG1z0NYHA4/3ZpiZF6ZSGPS8b6ewtr6xuVXcLu3s7u0flA+PmibJNOMNlshEt0NquBSKN1Cg5O1UcxqHkrfC0d3Mbz1xbUSiHnCc8iCmAyUiwShaqY3klvju1WWvXPFcbw6ySvycVCBHvVf+6vYTlsVcIZPUmI7vpRhMqEbBJJ+WupnhKWUjOuAdSxWNuQkm83un5MwqfRIl2pZCMld/T0xobMw4Dm1nTHFolr2Z+J/XyTC6CSZCpRlyxRaLokwSTMjsedIXmjOUY0so08LeStiQasrQRlSyIfjLL6+S5oXre65/f1mp1vI4inACp3AOPlxDFWpQhwYwkPAMr/DmPDovzrvzsWgtOPnMMfyB8/kD19+OiQ==</latexit><latexit sha1_base64="6q819vkUgiYHlbL7dyf4a4CC5DQ=">AAAB73icbVBNS8NAEJ3Ur1q/qh69LBbBU0hE0YtQ8NJjBfsBbSib7aZdutnE3YlQSv+EFw+KePXvePPfuG1z0NYHA4/3ZpiZF6ZSGPS8b6ewtr6xuVXcLu3s7u0flA+PmibJNOMNlshEt0NquBSKN1Cg5O1UcxqHkrfC0d3Mbz1xbUSiHnCc8iCmAyUiwShaqY3klvju1WWvXPFcbw6ySvycVCBHvVf+6vYTlsVcIZPUmI7vpRhMqEbBJJ+WupnhKWUjOuAdSxWNuQkm83un5MwqfRIl2pZCMld/T0xobMw4Dm1nTHFolr2Z+J/XyTC6CSZCpRlyxRaLokwSTMjsedIXmjOUY0so08LeStiQasrQRlSyIfjLL6+S5oXre65/f1mp1vI4inACp3AOPlxDFWpQhwYwkPAMr/DmPDovzrvzsWgtOPnMMfyB8/kD19+OiQ==</latexit><latexit sha1_base64="6q819vkUgiYHlbL7dyf4a4CC5DQ=">AAAB73icbVBNS8NAEJ3Ur1q/qh69LBbBU0hE0YtQ8NJjBfsBbSib7aZdutnE3YlQSv+EFw+KePXvePPfuG1z0NYHA4/3ZpiZF6ZSGPS8b6ewtr6xuVXcLu3s7u0flA+PmibJNOMNlshEt0NquBSKN1Cg5O1UcxqHkrfC0d3Mbz1xbUSiHnCc8iCmAyUiwShaqY3klvju1WWvXPFcbw6ySvycVCBHvVf+6vYTlsVcIZPUmI7vpRhMqEbBJJ+WupnhKWUjOuAdSxWNuQkm83un5MwqfRIl2pZCMld/T0xobMw4Dm1nTHFolr2Z+J/XyTC6CSZCpRlyxRaLokwSTMjsedIXmjOUY0so08LeStiQasrQRlSyIfjLL6+S5oXre65/f1mp1vI4inACp3AOPlxDFWpQhwYwkPAMr/DmPDovzrvzsWgtOPnMMfyB8/kD19+OiQ==</latexit><latexit sha1_base64="hP+6LrUf2d3tZaldqaQQvEKMXyw=">AAAB2XicbZDNSgMxFIXv1L86Vq1rN8EiuCozbnQpuHFZwbZCO5RM5k4bmskMyR2hDH0BF25EfC93vo3pz0JbDwQ+zknIvSculLQUBN9ebWd3b/+gfugfNfzjk9Nmo2fz0gjsilzl5jnmFpXU2CVJCp8LgzyLFfbj6f0i77+gsTLXTzQrMMr4WMtUCk7O6oyaraAdLMW2IVxDC9YaNb+GSS7KDDUJxa0dhEFBUcUNSaFw7g9LiwUXUz7GgUPNM7RRtRxzzi6dk7A0N+5oYkv394uKZ9bOstjdzDhN7Ga2MP/LBiWlt1EldVESarH6KC0Vo5wtdmaJNChIzRxwYaSblYkJN1yQa8Z3HYSbG29D77odBu3wMYA6nMMFXEEIN3AHD9CBLghI4BXevYn35n2suqp569LO4I+8zx84xIo4</latexit><latexit sha1_base64="Tx3si7JGLUTkXBb9gPKHBmqwXlQ=">AAAB5HicbZBLSwMxFIXv+Ky1anXrJlgEV0NGFN0IghuXFewD2qFk0kwbmsmMyR2hDP0Tblwo4m9y578xfSy09UDg45yE3HuiTEmLlH57a+sbm1vbpZ3ybmVv/6B6WGnaNDdcNHiqUtOOmBVKatFAiUq0MyNYEinRikZ307z1LIyVqX7EcSbChA20jCVn6Kw2khsS+JcXvWqN+nQmsgrBAmqwUL1X/er2U54nQiNXzNpOQDMMC2ZQciUm5W5uRcb4iA1Ex6FmibBhMZt3Qk6d0ydxatzRSGbu7xcFS6wdJ5G7mTAc2uVsav6XdXKMr8NC6ixHofn8ozhXBFMyXZ70pREc1dgB40a6WQkfMsM4uorKroRgeeVVaJ77AfWDBwolOIYTOIMAruAW7qEODeCg4AXe4N178l69j3lda96ityP4I+/zB7HgjSk=</latexit><latexit sha1_base64="Tx3si7JGLUTkXBb9gPKHBmqwXlQ=">AAAB5HicbZBLSwMxFIXv+Ky1anXrJlgEV0NGFN0IghuXFewD2qFk0kwbmsmMyR2hDP0Tblwo4m9y578xfSy09UDg45yE3HuiTEmLlH57a+sbm1vbpZ3ybmVv/6B6WGnaNDdcNHiqUtOOmBVKatFAiUq0MyNYEinRikZ307z1LIyVqX7EcSbChA20jCVn6Kw2khsS+JcXvWqN+nQmsgrBAmqwUL1X/er2U54nQiNXzNpOQDMMC2ZQciUm5W5uRcb4iA1Ex6FmibBhMZt3Qk6d0ydxatzRSGbu7xcFS6wdJ5G7mTAc2uVsav6XdXKMr8NC6ixHofn8ozhXBFMyXZ70pREc1dgB40a6WQkfMsM4uorKroRgeeVVaJ77AfWDBwolOIYTOIMAruAW7qEODeCg4AXe4N178l69j3lda96ityP4I+/zB7HgjSk=</latexit><latexit sha1_base64="1LiVteQRsPztfHLyLh3pSCPRNaY=">AAAB73icbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0YtQ8NJjBfsBbSib7aZdutnE3YlQQv+EFw+KePXvePPfuG1z0NYHA4/3ZpiZFyRSGHTdb6ewtr6xuVXcLu3s7u0flA+PWiZONeNNFstYdwJquBSKN1Gg5J1EcxoFkreD8d3Mbz9xbUSsHnCScD+iQyVCwShaqYPklnjVq8t+ueJW3TnIKvFyUoEcjX75qzeIWRpxhUxSY7qem6CfUY2CST4t9VLDE8rGdMi7lioaceNn83un5MwqAxLG2pZCMld/T2Q0MmYSBbYzojgyy95M/M/rphje+JlQSYpcscWiMJUEYzJ7ngyE5gzlxBLKtLC3EjaimjK0EZVsCN7yy6ukdVH13Kp371Zq9TyOIpzAKZyDB9dQgzo0oAkMJDzDK7w5j86L8+58LFoLTj5zDH/gfP4A1p+OhQ==</latexit><latexit sha1_base64="6q819vkUgiYHlbL7dyf4a4CC5DQ=">AAAB73icbVBNS8NAEJ3Ur1q/qh69LBbBU0hE0YtQ8NJjBfsBbSib7aZdutnE3YlQSv+EFw+KePXvePPfuG1z0NYHA4/3ZpiZF6ZSGPS8b6ewtr6xuVXcLu3s7u0flA+PmibJNOMNlshEt0NquBSKN1Cg5O1UcxqHkrfC0d3Mbz1xbUSiHnCc8iCmAyUiwShaqY3klvju1WWvXPFcbw6ySvycVCBHvVf+6vYTlsVcIZPUmI7vpRhMqEbBJJ+WupnhKWUjOuAdSxWNuQkm83un5MwqfRIl2pZCMld/T0xobMw4Dm1nTHFolr2Z+J/XyTC6CSZCpRlyxRaLokwSTMjsedIXmjOUY0so08LeStiQasrQRlSyIfjLL6+S5oXre65/f1mp1vI4inACp3AOPlxDFWpQhwYwkPAMr/DmPDovzrvzsWgtOPnMMfyB8/kD19+OiQ==</latexit><latexit sha1_base64="6q819vkUgiYHlbL7dyf4a4CC5DQ=">AAAB73icbVBNS8NAEJ3Ur1q/qh69LBbBU0hE0YtQ8NJjBfsBbSib7aZdutnE3YlQSv+EFw+KePXvePPfuG1z0NYHA4/3ZpiZF6ZSGPS8b6ewtr6xuVXcLu3s7u0flA+PmibJNOMNlshEt0NquBSKN1Cg5O1UcxqHkrfC0d3Mbz1xbUSiHnCc8iCmAyUiwShaqY3klvju1WWvXPFcbw6ySvycVCBHvVf+6vYTlsVcIZPUmI7vpRhMqEbBJJ+WupnhKWUjOuAdSxWNuQkm83un5MwqfRIl2pZCMld/T0xobMw4Dm1nTHFolr2Z+J/XyTC6CSZCpRlyxRaLokwSTMjsedIXmjOUY0so08LeStiQasrQRlSyIfjLL6+S5oXre65/f1mp1vI4inACp3AOPlxDFWpQhwYwkPAMr/DmPDovzrvzsWgtOPnMMfyB8/kD19+OiQ==</latexit><latexit sha1_base64="6q819vkUgiYHlbL7dyf4a4CC5DQ=">AAAB73icbVBNS8NAEJ3Ur1q/qh69LBbBU0hE0YtQ8NJjBfsBbSib7aZdutnE3YlQSv+EFw+KePXvePPfuG1z0NYHA4/3ZpiZF6ZSGPS8b6ewtr6xuVXcLu3s7u0flA+PmibJNOMNlshEt0NquBSKN1Cg5O1UcxqHkrfC0d3Mbz1xbUSiHnCc8iCmAyUiwShaqY3klvju1WWvXPFcbw6ySvycVCBHvVf+6vYTlsVcIZPUmI7vpRhMqEbBJJ+WupnhKWUjOuAdSxWNuQkm83un5MwqfRIl2pZCMld/T0xobMw4Dm1nTHFolr2Z+J/XyTC6CSZCpRlyxRaLokwSTMjsedIXmjOUY0so08LeStiQasrQRlSyIfjLL6+S5oXre65/f1mp1vI4inACp3AOPlxDFWpQhwYwkPAMr/DmPDovzrvzsWgtOPnMMfyB8/kD19+OiQ==</latexit><latexit sha1_base64="6q819vkUgiYHlbL7dyf4a4CC5DQ=">AAAB73icbVBNS8NAEJ3Ur1q/qh69LBbBU0hE0YtQ8NJjBfsBbSib7aZdutnE3YlQSv+EFw+KePXvePPfuG1z0NYHA4/3ZpiZF6ZSGPS8b6ewtr6xuVXcLu3s7u0flA+PmibJNOMNlshEt0NquBSKN1Cg5O1UcxqHkrfC0d3Mbz1xbUSiHnCc8iCmAyUiwShaqY3klvju1WWvXPFcbw6ySvycVCBHvVf+6vYTlsVcIZPUmI7vpRhMqEbBJJ+WupnhKWUjOuAdSxWNuQkm83un5MwqfRIl2pZCMld/T0xobMw4Dm1nTHFolr2Z+J/XyTC6CSZCpRlyxRaLokwSTMjsedIXmjOUY0so08LeStiQasrQRlSyIfjLL6+S5oXre65/f1mp1vI4inACp3AOPlxDFWpQhwYwkPAMr/DmPDovzrvzsWgtOPnMMfyB8/kD19+OiQ==</latexit><latexit sha1_base64="6q819vkUgiYHlbL7dyf4a4CC5DQ=">AAAB73icbVBNS8NAEJ3Ur1q/qh69LBbBU0hE0YtQ8NJjBfsBbSib7aZdutnE3YlQSv+EFw+KePXvePPfuG1z0NYHA4/3ZpiZF6ZSGPS8b6ewtr6xuVXcLu3s7u0flA+PmibJNOMNlshEt0NquBSKN1Cg5O1UcxqHkrfC0d3Mbz1xbUSiHnCc8iCmAyUiwShaqY3klvju1WWvXPFcbw6ySvycVCBHvVf+6vYTlsVcIZPUmI7vpRhMqEbBJJ+WupnhKWUjOuAdSxWNuQkm83un5MwqfRIl2pZCMld/T0xobMw4Dm1nTHFolr2Z+J/XyTC6CSZCpRlyxRaLokwSTMjsedIXmjOUY0so08LeStiQasrQRlSyIfjLL6+S5oXre65/f1mp1vI4inACp3AOPlxDFWpQhwYwkPAMr/DmPDovzrvzsWgtOPnMMfyB8/kD19+OiQ==</latexit><latexit sha1_base64="6q819vkUgiYHlbL7dyf4a4CC5DQ=">AAAB73icbVBNS8NAEJ3Ur1q/qh69LBbBU0hE0YtQ8NJjBfsBbSib7aZdutnE3YlQSv+EFw+KePXvePPfuG1z0NYHA4/3ZpiZF6ZSGPS8b6ewtr6xuVXcLu3s7u0flA+PmibJNOMNlshEt0NquBSKN1Cg5O1UcxqHkrfC0d3Mbz1xbUSiHnCc8iCmAyUiwShaqY3klvju1WWvXPFcbw6ySvycVCBHvVf+6vYTlsVcIZPUmI7vpRhMqEbBJJ+WupnhKWUjOuAdSxWNuQkm83un5MwqfRIl2pZCMld/T0xobMw4Dm1nTHFolr2Z+J/XyTC6CSZCpRlyxRaLokwSTMjsedIXmjOUY0so08LeStiQasrQRlSyIfjLL6+S5oXre65/f1mp1vI4inACp3AOPlxDFWpQhwYwkPAMr/DmPDovzrvzsWgtOPnMMfyB8/kD19+OiQ==</latexit>
Examples of exchangeability:Single group average
----+++++-
Second flip
Examples of exchangeability:Single group average
++++++++++
----+++++-
+-+--++-++
Etc …
…
Examples of exchangeability:Single group average
• Symmetric errors
• Errors independent
• Subjects drawn from a single population
And the assumptions are:
Examples of exchangeability:Two groups paired
Here we can only exchange scans within
each subject. I.e. Input 1 for Input 2, Input 3 for
Input 4 etc
Assumed covariance matrix
The implicit assumption here is that data from all subjects have the same uncertainty
and that there is no dependence between subjects
Allowed swap
Examples of exchangeability:Two groups paired
Assumed covariance matrix
The implicit assumption here is that data from all subjects have the same uncertainty
and that there is no dependence between subjects
Disallowed swap
Examples of exchangeability:Two groups paired
123456789
10
Original
21345687910
Perm 1
12346587910
Perm 2 …
Examples of exchangeability:Two groups paired
Examples of exchangeability:blocked ANOVA
Same as previous: We can only swap labels within
each subject
Assumed covariance matrix
Assumptions: All subjects from the same “population”,
no dependence between subjects and “compound
symmetry” within subjects
Allowed swap
Examples of exchangeability:blocked ANOVA
Assumed covariance matrix
Assumptions: All subjects from the same “population”,
no dependence between subjects and “compound
symmetry” within subjects
No longer allowed swap
Examples of exchangeability:blocked ANOVA
Assumed covariance matrix
Assumptions: All subjects from the same “population”,
no dependence between subjects and “compound
symmetry” within subjects
No longer allowed swap
Examples of exchangeability:blocked ANOVA
My advice: Keep it simple!Each subject
scanned like this
Taking 4 contrasts to 2nd level
We want to find areas that respond “linearly” to pain.
My advice: Keep it simple!
Each subject scanned like this
Taking 4 contrasts to 2nd level
Repeating this for four subjects
My advice: Keep it simple!
You have to assume this covariance matrix
And figure out this contrast
{ <latexit sha1_base64="RG03HEwJX1WBjc+2nWlsTV1XmgQ=">AAAB6XicbVBNS8NAEJ3Ur1q/oh69LBbBU0lE0GPRi8cq9gPaUDbbSbt0swm7G6GE/gMvHhTx6j/y5r9x2+agrQ8GHu/NMDMvTAXXxvO+ndLa+sbmVnm7srO7t3/gHh61dJIphk2WiER1QqpRcIlNw43ATqqQxqHAdji+nfntJ1SaJ/LRTFIMYjqUPOKMGis99PK+W/Vq3hxklfgFqUKBRt/96g0SlsUoDRNU667vpSbIqTKcCZxWepnGlLIxHWLXUklj1EE+v3RKzqwyIFGibElD5urviZzGWk/i0HbG1Iz0sjcT//O6mYmug5zLNDMo2WJRlAliEjJ7mwy4QmbExBLKFLe3EjaiijJjw6nYEPzll1dJ66LmezX//rJavyniKMMJnMI5+HAFdbiDBjSBQQTP8Apvzth5cd6dj0VrySlmjuEPnM8fm4SNZQ==</latexit><latexit sha1_base64="RG03HEwJX1WBjc+2nWlsTV1XmgQ=">AAAB6XicbVBNS8NAEJ3Ur1q/oh69LBbBU0lE0GPRi8cq9gPaUDbbSbt0swm7G6GE/gMvHhTx6j/y5r9x2+agrQ8GHu/NMDMvTAXXxvO+ndLa+sbmVnm7srO7t3/gHh61dJIphk2WiER1QqpRcIlNw43ATqqQxqHAdji+nfntJ1SaJ/LRTFIMYjqUPOKMGis99PK+W/Vq3hxklfgFqUKBRt/96g0SlsUoDRNU667vpSbIqTKcCZxWepnGlLIxHWLXUklj1EE+v3RKzqwyIFGibElD5urviZzGWk/i0HbG1Iz0sjcT//O6mYmug5zLNDMo2WJRlAliEjJ7mwy4QmbExBLKFLe3EjaiijJjw6nYEPzll1dJ66LmezX//rJavyniKMMJnMI5+HAFdbiDBjSBQQTP8Apvzth5cd6dj0VrySlmjuEPnM8fm4SNZQ==</latexit><latexit sha1_base64="RG03HEwJX1WBjc+2nWlsTV1XmgQ=">AAAB6XicbVBNS8NAEJ3Ur1q/oh69LBbBU0lE0GPRi8cq9gPaUDbbSbt0swm7G6GE/gMvHhTx6j/y5r9x2+agrQ8GHu/NMDMvTAXXxvO+ndLa+sbmVnm7srO7t3/gHh61dJIphk2WiER1QqpRcIlNw43ATqqQxqHAdji+nfntJ1SaJ/LRTFIMYjqUPOKMGis99PK+W/Vq3hxklfgFqUKBRt/96g0SlsUoDRNU667vpSbIqTKcCZxWepnGlLIxHWLXUklj1EE+v3RKzqwyIFGibElD5urviZzGWk/i0HbG1Iz0sjcT//O6mYmug5zLNDMo2WJRlAliEjJ7mwy4QmbExBLKFLe3EjaiijJjw6nYEPzll1dJ66LmezX//rJavyniKMMJnMI5+HAFdbiDBjSBQQTP8Apvzth5cd6dj0VrySlmjuEPnM8fm4SNZQ==</latexit><latexit sha1_base64="RG03HEwJX1WBjc+2nWlsTV1XmgQ=">AAAB6XicbVBNS8NAEJ3Ur1q/oh69LBbBU0lE0GPRi8cq9gPaUDbbSbt0swm7G6GE/gMvHhTx6j/y5r9x2+agrQ8GHu/NMDMvTAXXxvO+ndLa+sbmVnm7srO7t3/gHh61dJIphk2WiER1QqpRcIlNw43ATqqQxqHAdji+nfntJ1SaJ/LRTFIMYjqUPOKMGis99PK+W/Vq3hxklfgFqUKBRt/96g0SlsUoDRNU667vpSbIqTKcCZxWepnGlLIxHWLXUklj1EE+v3RKzqwyIFGibElD5urviZzGWk/i0HbG1Iz0sjcT//O6mYmug5zLNDMo2WJRlAliEjJ7mwy4QmbExBLKFLe3EjaiijJjw6nYEPzll1dJ66LmezX//rJavyniKMMJnMI5+HAFdbiDBjSBQQTP8Apvzth5cd6dj0VrySlmjuEPnM8fm4SNZQ==</latexit>
Why put yourself through all that pain?
My advice: Keep it simple!
Much nicer, no?When you can take a single contrast from
the first level
And get this at the second level
Assuming only symmetric errors
Warning pertaining to FSL 6.0.1
Do not use the Model setup wizard together with
Randomise in FSL 6.0.1