SAM: Significance Analysis of Microarrays
Parameter Information
SAM Models
Several experimental design models can be processed by SAM to provide results
based on the selected design. Each design is represented as a tab
in the SAM initialization dialog. Comprehensive coverage of the use of SAM
and analysis of various experimental designs are covered in the TMeV manual
and the reference list found in the 'references' selection of the 'help' menu of the
main TMeV tool bar.
Two-class unpaired
This model provides selection buttons to indicate which experiments should be
in group A, group B, or Neither group. Options exist (buttons below group selection
controls) to save a grouping scheme, load a grouping scheme, or reset the grouping.
Two-class paired
This model provides selection buttons to indicate which experiments should be
paired. Experiments which are paired are considered to be related observations,
for instance, samples from a particular subject before and after drug treatment.
Each experiment can be paired with only one other experiment. Once a pairing has been
made the pair is put into the pairings list and selection of the two experiments
is disabled. Pairings may be removed from the pairing list and reassigned.
Options exist (buttons below group selection
controls) to save a pairing scheme, load a pairing scheme, or reset the pairings.
Multi-class
This model requires one to initially specify an number of classes (>2).
Once that is done selection buttons allow the assignment of experiments
to the appropriate classes. Groupings can be saved, loaded, and reset using the
buttons below the group selection controls.
Censored survival
The selection controls for this model allow the inclusion or exclusion of particular experiments using
the check boxes to the left of the experiment names. For those experiments included
a corresponding time and a state is selected (Censored, or Dead).
Number of Permutations
This integer number indicates the number of times each vector should be permuted
and have a d-statistic computed.
Imputation Engine
SAM handles missing data by constructing or imputing
missing values using one of two available methods.
K-nearest neighbor imputer
This option imputes a value based on the k nearest neighbors (using Euclidean distance)
to the vector which is missing the value. The k nearest neighbors are selected
such that each should have a value for the missing vector element.
Row average imputer
This option imputes by taking the mean of the other elements in the vector which
is missing the value.
Hierarchical Clustering
This check box selects whether to perform hierarchical clustering on the elements in each cluster
created.