Which statistical test to compare samples in Violin Plot?
1
0
Entering edit mode
5.4 years ago
Ankit ▴ 500

Hi,

I have 5 samples (1 control +2 patients + 2 parents) for which I plotted an average methylation profile using violin plot. I would like to know which statistical test I need to apply to compare control vs patients 1/ patient2? An example plot is like similar to this google image:

Thanks

violin plots statistics methylation • 15k views
ADD COMMENT
1
Entering edit mode

The test to apply isn't determined by the type of plot but by the nature of the data and assumptions that can be made about its distribution.

ADD REPLY
0
Entering edit mode

Yes you are right. So as #Carambakaracho suggested, I will apply t-test?

ADD REPLY
0
Entering edit mode

That can't be decided without knowing the data.

ADD REPLY
0
Entering edit mode
5.4 years ago
Carambakaracho ★ 3.3k

To me a violin plot is mostly a fancy representation of a box-plot. And a box plot is (for gaussian distributions) a graphical representation of a t-test.

ADD COMMENT
0
Entering edit mode

Thanks for reply. So I apply t-test for control vs patients? Is it Unpaired t-test? Please help Thanks

ADD REPLY
1
Entering edit mode

You can't apply a t-test blindly. You need to make sure that the assumptions of the test are met so that it is valid.

ADD REPLY
0
Entering edit mode

So my data has ~750,000 methylation array probes.

I first plotted the boxplot and then violin.

I have five individuals with varying methylation profile for each probes. So which test do you think is applicable to this kind of data, if I want to compare control with any of the 2 patient I have?

Does my data full-fill the assumptions as mentioned here? or how will I test the assumptions are made?

ADD REPLY
0
Entering edit mode

Does my data full-fill the assumptions as mentioned here?

You don't show the plots you produced. What Jean-Karime wrote above still holds true.

That can't be decided without knowing the data.

ADD REPLY
0
Entering edit mode

We have methylation EPIC array data 850K for five individals as I previously mentioned.

After normalization and filtering ~750K probes left. I would like to compare global or overall methylation profle for control vs patient. I would like to see if there is a significant difference between the two considering all the probes.
Please suggest me how else I can better explain the data?

Are there any plots or values which I need to show here?

ADD REPLY
0
Entering edit mode

Are there any plots or values which I need to show here?

in case you want someone to give qualified answers, that might help. How about the violin plots and box plots you mentioned? With box plots histograms are helpful, too (though included in the violin plots).

ADD REPLY
0
Entering edit mode

![Here is the image attached]

https://ibb.co/g4BkhTY

ADD REPLY
0
Entering edit mode

I am still not clear which test to use. Please help.

ADD REPLY
0
Entering edit mode

HI Ankit, i'm not a specialist on methylation data (not even a bit).

Do you really think a simple test is what you want? I mean, you obviously have a difference between patient and control and not so much if any between parents and control. Would it be more interesting to figure where there's differences in methylation? What have others done with similar data? Here's a nature article with a linear regression analysis, something that feels far more appropriate.

But again, not my area of expertise...

ADD REPLY

Login before adding your answer.

Traffic: 1512 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6