ANOVA

From Informatics

Jump to: navigation, search

Contents

What is ANOVA

ANOVA stands for Analysis of Variance

It compares groups means and see if there's statistical significance between group means.

How we implement it

We implement ANOVA analysis in geWorkbench by using the ANOVA component in Tigr's MeV.

  • We wrap our parameters into MeV's format and call their method in their jar file.
  • After we got result from that method, we convert their result format back to our format.
  • So their code stay untouched (means we didn't modify any of their code).
  • Even for ANOVA grid version, we do the same way, so we don't need to maintain ANOVA code in two different places, we only need to copy the jar file into the lib dir

How to compare the result in our ANOVA with MeV's result (in order to make sure our result is correct)

  1. Use R-RMA-data.txt as example (the one you gave me but I add a header to make it valid)
  2. in MeV - Multiple Array Viewer - File - Load Data - Tab Delimited (TDMS) - select the file - click on upper left most cell with expression value - Load.
  3. click ANOVA icon in MAV (not the 2-Fact ANOVA), 3 groups, assign 3 groups. Change p-value threshold to 0.05 (as in geWorkbench, default is 0.05, but in MeV is 0.01), click analyze.
  4. On your left hand side (their project tree) in MAV, click on One-way ANOVA, click on Table Views, click on Significant Genes. You'll get a table similar to ours. (with different orders)
  5. Click on the header called Adj. p value (because in geworkbench, we only show adjusted p-value, so the p-value we ordered by, is actually their adj p-value)
  6. Now you use ANOVA in geworkbench as described in use case, (using "RMA Express, Tab-Delimited" file filter to load R-RMA-data.txt)
  7. Now you should be able to see the order of marker name in both geworkbench and MeV are the same, as well as our F-statistic equals their F ratio, our P-Value equals to their Adj. p value, etc.

Use Case

https://sharepoint.cu-genome.org/c2b2/Use%20Cases/ANOVA.doc

How to put ANOVA into geworkbench

put following in all.xml for analysis itself

   <plugin id="anova" name="Anova Analysis" class="org.geworkbench.components.anova.AnovaAnalysis" source="anova"/>

for visualize ANOVA result

   <plugin id="tabularDataViewer" name="Tabular Viewer" class="org.geworkbench.components.anova.gui.TabularDataViewer" source="anova">
       <gui-area name="VisualArea"/>
   </plugin>		
   <plugin id="colorMosaicPanel" name="Color Mosaic" class="org.geworkbench.components.colormosaic.ColorMosaicPanel" source="colormosaic">
       <gui-area name="VisualArea"/>
   </plugin>

How to run ANOVA

data

You'll need a microarray set, include at least 6 microarrays, so you can have three groups with at least two arrays in each groups.

web100.exp will work.

prerequests

In order for ANOVA to run, you'll need:

  • At least three groups of arrays selected and activated in geWorkbench.
  • Each group needs at least two arrays.

parameters

system test

System test 1.6 for Anova

How to put ANOVA into eclipse

You can check out source code for Anova from CVS server using following information:

  • Connection Type: extssh
  • Host: cbiocvs2.nci.nih.gov
  • Path: /share/content/gforge/geworkbench
  • Module: geworkbench/components/anova

Anova grid service

Anova grid analysis is written using CaGrid services and been deployed on cagridnode and afdev2.

To setup environment for debugging Anova grid service in eclipse, you'll need to obtain source code using following CVS information.

  • Connection type: extssh
  • Host: adcvs.cu-genome.org
  • Path: /cvs/magnet
  • Module: cagrid/cagrid_1_1/Anova

Then, you can use Build-server.xml to deploy the service in to your Tomcat.

Setup your MyEclipse to start the Tomcat for you in debug mode.

Now you can set break points in server side code (for Anova, usually in AnovaImpl.java or AnovaClient.java), and do the debugging.

XML model for Anova

We have XML model for Anova built using EA, currently the file path is file://caldev/Mark/Anova/

Includes, eap file, xmi file and Java Docs for APIs.

xmi file is an annotated Cagrid service model.

Bugs

Known bugs: Anova related bugs on Mantis

JUnit for Anova

AnovaAnalysisTest.java and AnovaAnalysisGuiTest.java

Personal tools