Ch11 - Statistical Analysis II: Advanced Approaches Flashcards
Hypothesis-driven analysis
Evaluation of data based on statistical tests of validity of a null hypothesis
multiple regression
statistical approaches that evaluate the relative contributions of several independent variables to a dependent variable
data-driven analysis
drawing inferences based on examination of intrinsic structure of data
component (data-driven analysis)
feature of a data set that represents some aspect of its intrinsic structure
data reduction
simplification of a data set by reducing its variability e.g, eliminating irrelevant, redundant, nonpredictive variables
principle component analysis (PCA)
common technique for data reduction
- simplifies high-dimensional data into smaller set of components that retains most of the variability of the original data set
orthogonal
property of 2 variables such that they are completely uncorrelated with each other
eigenimage
spacial maps
- generated by PCA that reflect orthogonal components of a complex image (time series of images)
eigenvector
set of values that describe component of the intrinsic variability in data
- following PCA
eigenvalue
mathematical describtion of amount of variability in data set that is accounted for by a given component
independent component analysis (ICA)
important class of data-driven analyses that identify stationary se of voxels whoses BOLD time course vary together over time and maximally distinguishable from other sets
mixing matrix
statistical describtion of how set of hypothesized sources (e.g., components) combine to form the obsereved data
spacial ICA
form of independent component analysis that generates components that have minimal spacial redundancy
temporal ICA
form of independent component analysis that generates components that have minimal temporal redundancy
resting-state connectivity
functional connectivity of a given brain region when measured while the participant is not performing any coordinated, purposeful task
partial least squares (PLS)
approach to analyzing functional neuroimaging data that is used to identify components whose amplitude is influenced by an experimental manipulation
latent variable
variable whose value is not directly measured but is inferred based on the values of other variables
permutation (significance testing)
approach that involve resampling original data to determine the size of and effect that might be obsereved with a given alpha level
hyperscanning
simultaneous collection of fMRI data from 2+ subjects who are interacting in an experimental paradigm
intersubject correlation
functional MRI time course that are shared by different individuals while performing the same experimental task/stimuli
reverse interference
reasoning from the outcome of a dependent variable to infer the state of a dependent variable
fiber tracts
bundles of axons that convey signals, as action potentials, from one brain region to another
structural connectivity
pattern of structural connections between regions based on known axonal projections
coactivation
simultanous activation of 2+ brain regions within single experimental task
- doesn’t imply that regions are functionally connected!
double dissociation
demonstration that 2 experimental manipulations have different effects on 2 dependent variables
seed voxel
voxel chosen as a starting point for a connectivity analysis
default network
set of brain regions whose activation tends to decrease during the performance of active, engaging tasks, but to increase during conditions of resting and reflection
psychophysiological interaction (PPI)
statistical approach for identifying the effect of an experimental manipulation on the functional connectivity between 2 brain regions
epiphenomenal
secondary consequence of a causal chain of processes, but playing no causal role in the process of interest
structural equation modeling (SEM)
statistical approach for testing a model of the relationships among a set of independent and dependent variables
dynamic causal modeling (DCM)
statistical approach for testing meodels of the connectivity between brain regions based on hypotheses about how experimental manipulations alter activation and connectivity between regions
Granger causality
A form of time series analysis that quantifies the information gained by using the past history of one variable to improve predictions of future values of another variable
Diffusion tensor imaging (DTI)
collection of images that provide information about the magnitude and direction of molecular diffusion
- for creation of fractional anisotropy
fractional anisotropy (FA)
preference for molecules to diffuse in an anisotropic manner
- FA value of 1 = diffusion along prefered axis
- FA value of 0 = diffusion similar in all directions
real-time analyses
set of computational steps designed for tge rapid analysis of fMRI data so that statistical tests can be conducted immediately following the acquisition of images
presurgical planning
use of fMRI or another neuroscience technique to map particular functions in a single individual to guide clinical decision-making about the potential course and consequences of neurosurgery in that individual
biofeedback
providing an explicit indicator of some physiological process (e.g. beating heart) so individual can attempt to regulate that activation/guide behavior
biomarker
phenotypic feature (physical, pysiological, behavioral) that provides robust predictor of some experimentally or clinically important outcome
subsequent memory
approach to fMRI analysis
- sorts experimental stimuli based on whether they were remembered or forgotten in later testing session
- allows identification of brain regions whose activation predicts successful encoding of stimulus
logistical regression
subset of regression analysis that uses set of independent variables to predict binary outcome variable
pattern classification
attemt to separate individual exemplars into different categories by constructing a set of decision rules based on some combination of their feature
E.g.:
machine learning
subdicipline within computer science that develops algorithmic rules for relating input data to desirable outputs
multi-voxel pattern analysis (MVPA)
approach for pattern classification in fMRI research that uses data of relative changes of activation across set of voxels as input
feature selection
initial step in pattern classification that involves determination of which input variables should be included in the classification algorithm
searchlight
approach to feature selection in pattern classification
- reflects geometrically defined ROI (e.g. sphere of 5 voxel radius) that can be moved through the brain
training set
in pattern classification analysis
- part of data set that is used to develop the classification algorithm
testing set
in pattern classification
- novel part of data set used to evaluate the robustness of the classification algorithm
support vector machines (SVMs)
class of algorithm used in pattern classification that attempts to identify the combination of features in the original data set that can most effectively differentiate between 2 categories
cross-validation
in pattern classification analysis
- approach to evaluate the effectiveness of classification using given feature set
- iterated generation and testing of classifiers based on different parts of the same training set
