MATLAB and Simulink resources for Arduino, LEGO, and Raspberry Pi

Learn moreOpportunities for recent engineering grads.

Apply TodayTo resolve issues starting MATLAB on Mac OS X 10.10 (Yosemite) visit: http://www.mathworks.com/matlabcentral/answers/159016

Asked by Mariam Sheha on 22 Jun 2013

Hello Everybody;

Hope you are all doing well...

Can anybody explain the main difference between f-test and t-test?,

where i want to use it as method to identify the effective features among a list 60 features extracted out of two types of data (i.e, data 1 composed of 30 image for each image 60 features are extracted/ data 2 composed of 40 images for each image 60 features are extracted), i do search but i still misunderstand the difference..

Note that: the matlab equations for t- test is (h = ttest2(x,y)) and f-test is (H = vartest2(X,Y))

Thanks Alot;

Answer by Shashank Prasanna on 22 Jun 2013

Accepted answer

t-test is used to test if two sample have the same mean. The assumptions are that they are samples from normal distribution.

f-test is used to test if two sample have the same variance. Same assumptions hold.

I have little to no experience in image processing to comment on if these tests make sense to your application. A little more info of the problem you are trying to solve will be useful.

If you are however solving a classification problem (categorizing your images) You can use stepwise logistic regression with F-statistics criterion to reduce your predictor dimension:

http://www.mathworks.com/help/stats/generalizedlinearmodel.stepwise.html

Alternatively you could use PCA as Image Analyst suggested which does not take into account the response when reducing the dimensionality.

Mariam Sheha on 22 Jun 2013

Thanks for Replaying...

i am not on the classification phase, just trying to judge the extracted features that i did extract from an image to be quit effective to insert it as an input data for my classifier...

anyway, i will try all the proposed suggestions to get the best results..

Thanks A lot;

Answer by Image Analyst on 22 Jun 2013

Were the Wikipedia explanations not understandable? Anyway, I'm no statistician but I think you'd want **Principal Components Analysis**, rather than t-test of F-test, if you want to figure out which of 60 features are the most important.

Mariam Sheha on 22 Jun 2013

Thanks a lot for replaying....

I don't say that 'i don't understand it,but the difference between using t-test or f-test upon my data *as listed above* was my question!

Anyway, i had partially get the answer, which specify that using t-test may be much more effective for my data than F-test, also using PCA to figure out the most effective features is a good idea...

Thanks a lot

Image Analyst on 22 Jun 2013

OK, whatever...I have no idea how you'd use either the t-test *or* F-test with 60 features to "judge the features", but if you do, then that's all that counts.

## 0 Comments