Maf and MIA Toolbox PCA Quick Start: Difference between pages

From Eigenvector Research Documentation Wiki
(Difference between pages)
Jump to navigation Jump to search
imported>Scott
No edit summary
 
imported>Scott
No edit summary
 
Line 1: Line 1:
===Purpose===
[[Loading Data Quick Start | Previous Topic: Loading Data]]
Maximum Autocorrelative Factors.
{| border="1" cellpadding="5" cellspacing="0" align="left"
===Synopsis===
|-
:[scores,loads,mn,ssq] = maf(x)
|valign="top" |
===Description===
[[Image:LoadExportSMBread.png |right |(Click to Enlarge)]]
This example uses the '''smbread''' dataset. Load '''smbread''' by clicking on '''Load Demo Data''' in the tree, then '''Swedish Knackerbrod'''. This will load the data into Image Manager. Click on the item in the Loaded Images section to view it and make it the current selected image. Next, click on the export to Analysis button. This will open the Analysis GUI and load '''smbread'''.
|-
|valign="top" |
[[Image:RecalcBreadImage4PCs.png |right| 500px |(Click to Enlarge)]]
The Analysis interface will open with '''smbread''' loaded. Next, select '''PCA''' from the '''Analysis''' menu and then click the '''Model''' icon in the Status Pane. By default a model with one PC will be calculated. Let's try 4 PCs by selecting row 4 and clicking the '''Model''' icon again.


NOTE: The function MAF is obsolete and will be removed in future versions. Please use [[maxautofactors]].
|-
|valign="top"|
[[Image:build_model.004.png |right| 500px |(Click to Enlarge)]]
The Analysis GUI will then appear with the given data already loaded. (The large '''X''' (block) button in the Analysis GUI will appear depressed indicating the data is loaded - passing the cursor over the button will provide a summary of the '''X''' block data.)
|-
|valign="top"|
[[Image:build_model.005.png |right| 500px |(Click to Enlarge)]]
While each analysis type has its own nuances, in general the steps to build a model are:


Similar to PCA, MAF uses singular value decomposition to provide a model of data (x) which captures maximum spatially-correlated variance. The resulting scores (scores), loadings (loads), and mean spectrum (mn) can be used to reconstruct the mean-centered data matrix X_mn:
# load the data
X_mn = scores\*loads
# view the data
The difference between PCA and MAF is that MAF extracts loadings which are highly correlated in the spatial dimension of an image. In addition, MAF always returns the entire set of components up to the rank of the data matrix.
# choose preprocessing
Input X is either an Image DataSet object or a three-way double matrix (first two dimensions are spatial, last is variable).
# select the type of cross-validation to use (if any is desired - see [[Using Cross-Validation]])
Output ssq is an experimental sum of squares captured table but because the nature of the decomposition, this table is only approximate. It includes the component number (column 1), estimated eigenvalue (column 2), and the estimated captured per variable and total (columns 3 and 4, respectively).
# build the model
===See Also===
# review the model
[[mcr]], [[parafac]], [[pca]]
 
There is more than one way to accomplish each of the above steps. For example, preprocessing can be set for the '''X''' block by
* clicking the '''"Choose Preprocessing"''' button in the Analysis Flowchart
* clicking the '''P''' button next to the '''X''' button
* select '''"Preprocess"''' menu item
 
('''Hint:''' Once selected, use the Preprocess/X-block and Preprocess/Y-block menus to save the current preprocessing as the default.)
 
You can quickly view your data by right-clicking on the appropriate button ('''X''' or '''Y''') and then select '''"Plot Data"'''
 
|-
|valign="top"|
[[Image:build_model.006.png|right|  |(Click to Enlarge)]]
Perhaps the quickest way for new users to complete the remaining steps to build the model is to use the flowchart.  
 
|}
[[Review Results Quick Start | Next Topic: Review Results]]

Revision as of 17:41, 9 October 2009

Previous Topic: Loading Data

(Click to Enlarge)

This example uses the smbread dataset. Load smbread by clicking on Load Demo Data in the tree, then Swedish Knackerbrod. This will load the data into Image Manager. Click on the item in the Loaded Images section to view it and make it the current selected image. Next, click on the export to Analysis button. This will open the Analysis GUI and load smbread.

(Click to Enlarge)

The Analysis interface will open with smbread loaded. Next, select PCA from the Analysis menu and then click the Model icon in the Status Pane. By default a model with one PC will be calculated. Let's try 4 PCs by selecting row 4 and clicking the Model icon again.

(Click to Enlarge)

The Analysis GUI will then appear with the given data already loaded. (The large X (block) button in the Analysis GUI will appear depressed indicating the data is loaded - passing the cursor over the button will provide a summary of the X block data.)

(Click to Enlarge)

While each analysis type has its own nuances, in general the steps to build a model are:

  1. load the data
  2. view the data
  3. choose preprocessing
  4. select the type of cross-validation to use (if any is desired - see Using Cross-Validation)
  5. build the model
  6. review the model

There is more than one way to accomplish each of the above steps. For example, preprocessing can be set for the X block by

  • clicking the "Choose Preprocessing" button in the Analysis Flowchart
  • clicking the P button next to the X button
  • select "Preprocess" menu item

(Hint: Once selected, use the Preprocess/X-block and Preprocess/Y-block menus to save the current preprocessing as the default.)

You can quickly view your data by right-clicking on the appropriate button (X or Y) and then select "Plot Data"

(Click to Enlarge)

Perhaps the quickest way for new users to complete the remaining steps to build the model is to use the flowchart.

Next Topic: Review Results

Retrieved from "https://www.wiki.eigenvector.com/index.php?title=Maf&oldid=5611"