FMRI Group Analysis GLM. Voxelwise group analysis. Design matrix. Subject groupings. Group effect size. statistics. Effect size


 Bernice Harrington
 1 years ago
 Views:
Transcription
1 FMRI Group Analysis Voxelwise group analysis Standardspace brain atlas Subject groupings Design matrix subjects Singlesubject Singlesubject effect size Singlesubject effect statistics size Singlesubject effect statistics size effect statistics size statistics Register subjects into a standard space Effect size subjectseries subjects GLM Group effect size statistics Contrast Statistic Image Significant voxels/clusters Effect size statistics Thresholding
2 MultiLevel FMRI analysis uses GLM at both lower and higher levels typically need to infer across multiple subjects, sometimes multiple groups and/or multiple sessions Group 1 Difference? Group 2 Mark Steve Karl Will Tom Andrew Josephine Anna Hanna Sebastian Lydia Elisabeth session 1 session 1 session 1 session 1 session 1 session 1 session 1 session 1 session 1 session 1 session 1 session 1 session 2 session 2 session 2 session 2 session 2 session 2 session 2 session 2 session 2 session 2 session 2 session 2 session 3 session 3 session 3 session 3 session 3 session 3 session 3 session 3 session 3 session 3 session 3 session 3 session 4 session 4 session 4 session 4 session 4 session 4 session 4 session 4 questions of interest involve comparisons at the highest level
3 A simple example Does the group activate on average? Group Mark Steve Karl Will Tom Andrew
4 A simple example Does the group activate on average? Group Mark Steve Karl Will Tom Andrew effect size
5 A simple example Does the group activate on average? Group Mark Steve Karl Will Tom Andrew Y k = X k k + k Firstlevel GLM on Mark s 4D FMRI data set effect size
6 A simple example Does the group activate on average? Group Mark Steve Karl Will Tom Andrew Y k = X k k + k Mark s effect size effect size
7 A simple example Does the group activate on average? Group Mark Steve Karl Will Tom Andrew Y k = X k k + k Mark s withinsubject variance effect size
8 A simple example Does the group activate on average? Group Mark Steve Karl Will Tom Andrew Y K = X K K + K All firstlevel GLMs on 6 FMRI data set effect size
9 A simple example Does the group activate on average? Group Mark Steve Karl Will Tom Andrew What group mean are we after? Is it: 1. The group mean for those exact 6 subjects? FixedEffects (FE) Analysis 2. The group mean for the population from which these 6 subjects were drawn? MixedEffects (ME) analysis
10 FixedEffects Analysis Do these exact 6 subjects activate on average? Group Mark Steve Karl Will Tom Andrew estimate group effect size as straightforward mean across lowerlevel estimates effect size 6 g = 1 6 k=1 k
11 FixedEffects Analysis Do these exact 6 subjects activate on average? Group Mark Steve Karl Will Tom Andrew Y K = X K K + K X g = K = X g g Group mean effect size g = k k=1
12 FixedEffects Analysis Do these exact 6 subjects activate on average? Group Mark Steve Karl Will Tom Andrew Y K = X K K + K K = X g Fixed Effects Analysis: Consider only these 6 subjects estimate the mean across these subject only variance is withinsubject variance g
13 A simple example Does the group activate on average? Group Mark Steve Karl Keith Tom Andrew What group mean are we after? Is it: 1. The group mean for those exact 6 subjects? FixedEffects (FE) Analysis 2. The group mean for the population from which these 6 subjects were drawn? MixedEffects (ME) analysis
14 MixedEffects Analysis Does the population activate on average? Group Mark Steve Karl Keith Tom Andrew Y K = X K K + K Consider the distribution over the population from which our 6 subjects were sampled: g effect size g k 2 g is the betweensubject variance
15 MixedEffects Analysis Does the population activate on average? Group Mark Steve Karl Keith Tom Andrew Y K = X K K + K X g = K = X g g + g Population mean betweensubject variation g effect size g k
16 MixedEffects Analysis Does the population activate on average? Group Mark Steve Karl Keith Tom Andrew Y K = X K K = X g K + K g + g MixedEffects Analysis: Consider the 6 subjects as samples from a wider population estimate the mean across the population betweensubject variance accounts for random sampling
17 AllinOne Approach Group 1 Difference? Group 2 Mark Steve Karl Will Tom Andrew Josephine Anna Hanna Sebastian Lydia Elisabeth session 1 session 1 session 1 session 1 session 1 session 1 session 1 session 1 session 1 session 1 session 1 session 1 session 2 session 2 session 2 session 2 session 2 session 2 session 2 session 2 session 2 session 2 session 2 session 2 session 3 session 3 session 3 session 3 session 3 session 3 session 3 session 3 session 3 session 3 session 3 session 3 session 4 session 4 session 4 session 4 session 4 session 4 session 4 session 4 Could use one (huge) GLM to infer group difference difficult to ask subquestions in isolation computationally demanding need to process again when new data is acquired
18 Summary Statistics Approach In FEAT estimate levels one stage at a time At each level: Inputs are summary stats from levels below (or FMRI data at the lowest level) Outputs are summary stats or statistic maps for inference Need to ensure formal equivalence between different approaches! Group difference Group Subject Session
19 FLAME FMRIB s Local Analysis of Mixed Effects Fully Bayesian framework use noncentral tdistributions: Input COPES, VARCOPES & DOFs from lowerlevel estimate COPES, VARCOPES & DOFs at current level pass these up Infer at top level Equivalent to AllinOne approach ZStats Group difference COPES VARCOPES DOFs COPES VARCOPES DOFs COPES VARCOPES DOFs Group Subject Session
20 FLAME Inference Default is: FLAME1: fast approximation for all voxels (using marginal variance MAP estimates) Optional slower, slightly more accurate approach: FLAME1+2: FLAME1 for all voxels, FLAME2 for voxels close to threshold FLAME2: MCMC sampling technique
21 Choosing Inference Approach 1. Fixed Effects Use for intermediate/top levels 2. Mixed Effects  OLS Use at top level: quick and less accurate 3. Mixed Effects  FLAME 1 Use at top level: less quick but more accurate 4. Mixed Effects  FLAME 1+2 Use at top level: slow but even more accurate
22 FLAME vs. OLS allow different withinlevel variances (e.g. patients vs. controls) pat ctl allow nonbalanced designs (e.g. containing behavioural scores) effect size... allow unequal group sizes solve the negative variance problem Session < < Subject Group
23 FLAME vs. OLS Two ways in which FLAME can give different Zstats compared to OLS: higher Z due to increased efficiency from using lowerlevel variance heterogeneity FLAME OLS
24 FLAME vs. OLS Two ways in which FLAME can give different Zstats compared to OLS: Lower Z due to higherlevel variance being constrained to be positive (i.e. solve the implied negative variance problem) FLAME OLS
25 Multiple Group Variances can deal with multiple variances separate variance will be estimated for each variance group (be aware of #observations for each estimate, though!) group effect size design matrices need to be separable, i.e. EVs only have nonzero values for a single group pat ctl valid invalid
26 Examples
27 Single Group Average We have 8 subjects  all in one group  and want the mean group average: Does the group activate on average? estimate mean estimate stderror (FE or ME) test significance of mean > subject effect size >?
28 Single Group Average Does the group activate on average? subject effect size
29 Single Group Average Does the group activate on average?
30 subject Unpaired TwoGroup Difference We have two groups (e.g. 9 patients, 7 controls) with different betweensubject variance Is there a significant group difference? estimate means estimate stderrors (FE or ME) test significance of difference in means >? effect size
31 subject Unpaired TwoGroup Difference Is there a significant group difference? effect size
32 subject Unpaired TwoGroup Difference Is there a significant group difference? effect size
33 Unpaired TwoGroup Difference Is there a significant group difference?
34 Paired TTest 8 subjects scanned under 2 conditions (A,B) Is there a significant difference between conditions? subject effect size
35 Paired TTest 8 subjects scanned under 2 conditions (A,B) Is there a significant difference between conditions? try nonpaired ttest subject >? effect size
36 Paired TTest 8 subjects scanned under 2 conditions (A,B) Is there a significant difference between conditions? data demeaned data subject subject effect size subject mean accounts for large prop. of the overall variance effect size
37 Paired TTest 8 subjects scanned under 2 conditions (A,B) Is there a significant difference between conditions? data demeaned data subject subject effect size subject mean accounts for large prop. of the overall variance >? effect size
38 Paired TTest Is there a significant difference between conditions? subject effect size
39 Paired TTest Is there a significant difference between conditions? subject effect size
40 Paired TTest Is there a significant difference between conditions? EV1models the AB paired difference; EVs 29 are confounds which model out each subject s mean
41 Paired TTest Is there a significant difference between conditions?
42 MultiSession & MultiSubject 5 subjects each have three sessions. Does the group activate on average? Use three levels: in the second level we model the withinsubject repeated measure
43 MultiSession & MultiSubject 5 subjects each have three sessions. Does the group activate on average? Use three levels: in the third level we model the betweensubjects variance
44 MultiSession & MultiSubject 5 subjects each have three sessions. Does the group activate on average? Use three levels: in the second level we model the within subject repeated measure typically using fixed effects(!) as #sessions are small in the third level we model the between subjects variance using fixed or mixed effects
45 Reducing variance Does the group activate on average? subject subject >? effect size mean effect size large relative to std error >? mean effect size small relative to std error effect size
46 Reducing variance Does the group activate on average? subject subject >? effect size mean effect size large relative to std error >? mean effect size large relative to std error effect size
47 Single Group Average & Covariates We have 7 subjects  all in one group. We also have additional measurements (e.g. age; disability score; behavioural measures like reaction times): use covariates to explain variation Does the group activate on average? estimate mean estimate stderror (FE or ME) subject effect size
48 Single Group Average & Covariates We have 7 subjects  all in one group. We also have additional measurements (e.g. age; disability score; behavioural measures like reaction times): use covariates to explain variation Does the group activate on average? estimate mean estimate stderror (FE or ME) subject effect size slow RT fast
49 Single Group Average & Covariates We have 7 subjects  all in one group. We also have additional measurements (e.g. age; disability score; behavioural measures like reaction times): use covariates to explain variation Does the group activate on average? estimate mean estimate stderror (FE or ME) subject effect size slow RT fast
50 Single Group Average & Covariates We have 7 subjects  all in one group. We also have additional measurements (e.g. age; disability score; behavioural measures like reaction times): use covariates to explain variation Does the group activate on average? estimate mean estimate stderror (FE or ME) subject effect size slow RT fast
51 Single Group Average & Covariates We have 7 subjects  all in one group. We also have additional measurements (e.g. age; disability score; behavioural measures like reaction times): use covariates to explain variation Does the group activate on average? estimate mean estimate stderror (FE or ME) subject effect size slow RT fast
52 Single Group Average & Covariates Does the group activate on average? use covariates to explain variation need to demean additional covariates!
53 FEAT Group Analysis Run FEAT on raw FMRI data to get firstlevel.feat directories, each one with several (consistent) COPEs lowres copen/varcopen.feat/stats when higherlevel FEAT is run, highres copen/ varcopen.feat/reg_standard
54 FEAT Group Analysis Run secondlevel FEAT to get one.gfeat directory Inputs can be lowerlevel.feat dirs or lowerlevel COPEs the secondlevel GLM analysis is run separately for each firstlevel COPE each lowerlevel COPE generates its own.feat directory inside the.gfeat dir
55 That s all folks
56 Appendix:
57 3 groups of subjects Group Ftests Is any of the groups activating on average?
58 ANOVA: 1factor 4levels 8 subjects, 1 factor at 4 levels Is there any effect? EV1 fits cond. D, EV2 fits cond A relative to D etc. Ftest shows any difference between levels
59 ANOVA: 2factor 2levels 8 subjects, 2 factor at 2 levels. FE Anova: 3 Ftests give standard results for factor A, B and interaction If both factors are random effects then Fa=fstat1/fstat3, Fb=fstat2/fstat3ME ME: if fixed fact. is A, Fa=fstat1/fstat3
60 ANOVA: 3factor 2levels 16 subjects, 3 factor at 2 levels. FixedEffects ANOVA: For random/mixed effects need different Fs.
61 Understanding FEAT dirs Firstlevel analysis:
62 Understanding FEAT dirs Secondlevel analysis:
63 That s all folks
FixedEffect Versus RandomEffects Models
CHAPTER 13 FixedEffect Versus RandomEffects Models Introduction Definition of a summary effect Estimating the summary effect Extreme effect size in a large study or a small study Confidence interval
More informationOneWay Analysis of Variance (ANOVA) Example Problem
OneWay Analysis of Variance (ANOVA) Example Problem Introduction Analysis of Variance (ANOVA) is a hypothesistesting technique used to test the equality of two or more population (or treatment) means
More informationANOVA ANOVA. TwoWay ANOVA. OneWay ANOVA. When to use ANOVA ANOVA. Analysis of Variance. Chapter 16. A procedure for comparing more than two groups
ANOVA ANOVA Analysis of Variance Chapter 6 A procedure for comparing more than two groups independent variable: smoking status nonsmoking one pack a day > two packs a day dependent variable: number of
More informationHow to Deal with The LanguageasFixedEffect Fallacy : Common Misconceptions and Alternative Solutions
Journal of Memory and Language 41, 416 46 (1999) Article ID jmla.1999.650, available online at http://www.idealibrary.com on How to Deal with The LanguageasFixedEffect Fallacy : Common Misconceptions
More informationThe InStat guide to choosing and interpreting statistical tests
Version 3.0 The InStat guide to choosing and interpreting statistical tests Harvey Motulsky 19902003, GraphPad Software, Inc. All rights reserved. Program design, manual and help screens: Programming:
More informationHighDimensional Image Warping
Chapter 4 HighDimensional Image Warping John Ashburner & Karl J. Friston The Wellcome Dept. of Imaging Neuroscience, 12 Queen Square, London WC1N 3BG, UK. Contents 4.1 Introduction.................................
More informationBeyond Baseline and Followup: The Case for More T in Experiments * David McKenzie, World Bank. Abstract
Beyond Baseline and Followup: The Case for More T in Experiments * David McKenzie, World Bank Abstract The vast majority of randomized experiments in economics rely on a single baseline and single followup
More informationSawtooth Software. How Many Questions Should You Ask in ChoiceBased Conjoint Studies? RESEARCH PAPER SERIES
Sawtooth Software RESEARCH PAPER SERIES How Many Questions Should You Ask in ChoiceBased Conjoint Studies? Richard M. Johnson and Bryan K. Orme, Sawtooth Software, Inc. 1996 Copyright 19962002, Sawtooth
More informationDesign of Experiments (DOE)
MINITAB ASSISTANT WHITE PAPER This paper explains the research conducted by Minitab statisticians to develop the methods and data checks used in the Assistant in Minitab 17 Statistical Software. Design
More informationResults from the 2014 AP Statistics Exam. Jessica Utts, University of California, Irvine Chief Reader, AP Statistics jutts@uci.edu
Results from the 2014 AP Statistics Exam Jessica Utts, University of California, Irvine Chief Reader, AP Statistics jutts@uci.edu The six freeresponse questions Question #1: Extracurricular activities
More informationWhen and why do people avoid unknown probabilities in decisions under uncertainty? Testing some predictions from optimal foraging theory
COGNITION Cognition 72 (1999) 269±304 www.elsevier.com/locate/cognit When and why do people avoid unknown probabilities in decisions under uncertainty? Testing some predictions from optimal foraging theory
More informationPRINCIPAL COMPONENT ANALYSIS
1 Chapter 1 PRINCIPAL COMPONENT ANALYSIS Introduction: The Basics of Principal Component Analysis........................... 2 A Variable Reduction Procedure.......................................... 2
More informationEffect Sizes. Null Hypothesis Significance Testing (NHST) C8057 (Research Methods 2): Effect Sizes
Effect Sizes Null Hypothesis Significance Testing (NHST) When you read an empirical paper, the first question you should ask is how important is the effect obtained. When carrying out research we collect
More informationTwoWay Independent ANOVA Using SPSS
TwoWay Independent ANOVA Using SPSS Introduction Up to now we have looked only at situations in which a single independent variable was manipulated. Now we move onto more complex designs in which more
More informationSizeConstrained Weighted Set Cover
SizeConstrained Weighted Set Cover Lukasz Golab 1, Flip Korn 2, Feng Li 3, arna Saha 4 and Divesh Srivastava 5 1 University of Waterloo, Canada, lgolab@uwaterloo.ca 2 Google Research, flip@google.com
More informationA Fast Learning Algorithm for Deep Belief Nets
LETTER Communicated by Yann Le Cun A Fast Learning Algorithm for Deep Belief Nets Geoffrey E. Hinton hinton@cs.toronto.edu Simon Osindero osindero@cs.toronto.edu Department of Computer Science, University
More informationTesting a Hypothesis about Two Independent Means
1314 Testing a Hypothesis about Two Independent Means How can you test the null hypothesis that two population means are equal, based on the results observed in two independent samples? Why can t you use
More informationNSSE MultiYear Data Analysis Guide
NSSE MultiYear Data Analysis Guide About This Guide Questions from NSSE users about the best approach to using results from multiple administrations are increasingly common. More than three quarters of
More informationI like it... I like it not: Evaluating User Ratings Noise in Recommender Systems
I like it... I like it not: Evaluating User Ratings Noise in Recommender Systems Xavier Amatriain, Josep M. Pujol, and Nuria Oliver Telefonica Research Recent growing interest in predicting and influencing
More informationMisunderstandings between experimentalists and observationalists about causal inference
J. R. Statist. Soc. A (2008) 171, Part 2, pp. 481 502 Misunderstandings between experimentalists and observationalists about causal inference Kosuke Imai, Princeton University, USA Gary King Harvard University,
More informationThe Bonferonni and Šidák Corrections for Multiple Comparisons
The Bonferonni and Šidák Corrections for Multiple Comparisons Hervé Abdi 1 1 Overview The more tests we perform on a set of data, the more likely we are to reject the null hypothesis when it is true (i.e.,
More informationScalable Collaborative Filtering with Jointly Derived Neighborhood Interpolation Weights
Seventh IEEE International Conference on Data Mining Scalable Collaborative Filtering with Jointly Derived Neighborhood Interpolation Weights Robert M. Bell and Yehuda Koren AT&T Labs Research 180 Park
More informationStudent Guide to SPSS Barnard College Department of Biological Sciences
Student Guide to SPSS Barnard College Department of Biological Sciences Dan Flynn Table of Contents Introduction... 2 Basics... 4 Starting SPSS... 4 Navigating... 4 Data Editor... 5 SPSS Viewer... 6 Getting
More information8 INTERPRETATION OF SURVEY RESULTS
8 INTERPRETATION OF SURVEY RESULTS 8.1 Introduction This chapter discusses the interpretation of survey results, primarily those of the final status survey. Interpreting a survey s results is most straightforward
More informationHandling Missing Values when Applying Classification Models
Journal of Machine Learning Research 8 (2007) 16251657 Submitted 7/05; Revised 5/06; Published 7/07 Handling Missing Values when Applying Classification Models Maytal SaarTsechansky The University of
More informationStatistical Comparisons of Classifiers over Multiple Data Sets
Journal of Machine Learning Research 7 (2006) 1 30 Submitted 8/04; Revised 4/05; Published 1/06 Statistical Comparisons of Classifiers over Multiple Data Sets Janez Demšar Faculty of Computer and Information
More informationEVALUATION OF GAUSSIAN PROCESSES AND OTHER METHODS FOR NONLINEAR REGRESSION. Carl Edward Rasmussen
EVALUATION OF GAUSSIAN PROCESSES AND OTHER METHODS FOR NONLINEAR REGRESSION Carl Edward Rasmussen A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy, Graduate
More informationThe Wondrous World of fmri statistics
Outline The Wondrous World of fmri statistics FMRI data and Statistics course, Leiden, 1132008 The General Linear Model Overview of fmri data analysis steps fmri timeseries Modeling effects of interest
More informationAprimary strategy for including. Extended Time as a Testing Accommodation: Its Effects and Perceived Consequences. Exceptional Children
Vol. 70, No.3, pp. 349367. 2004 Council for. Extended Time as a Testing Accommodation: Its Effects and Perceived Consequences STEPHEN N. ELLIOTT University of WisconsinMadison ANN M. MARQUART Heartland
More informationCOMMENTS AND CONTROVERSIES Why VoxelBased Morphometry Should Be Used
NeuroImage 14, 1238 1243 (2001) doi:10.1006/nimg.2001.0961, available online at http://www.idealibrary.com on COMMENTS AND CONTROVERSIES Why VoxelBased Morphometry Should Be Used John Ashburner 1 and
More information