Question: The selection of bandwidth of GSVA
0
gravatar for Xingyu
18 months ago by
Xingyu0
Xingyu0 wrote:

Hi,

I'm learning GSVA paper recently and I'm wondering why GSVA chose Si/4 as the bandwidth for a Gaussian kenel estimation. I have read all the document I can find for GSVA but still didn't find the answer. Could you please give me the reason of this choice?

Thanks!

gsva • 291 views
ADD COMMENTlink modified 18 months ago by Robert Castelo2.3k • written 18 months ago by Xingyu0
Answer: The selection of bandwidth of GSVA
0
gravatar for Robert Castelo
18 months ago by
Robert Castelo2.3k
Spain/Barcelona/Universitat Pompeu Fabra
Robert Castelo2.3k wrote:

hi,

we chose this fixed setting because from our experience it worked well in practice.

robert.

ADD COMMENTlink written 18 months ago by Robert Castelo2.3k

Hi Robert,

I wonder is there a quantitative metric for evaluating how well it worked or just chose one which gave best results?

Thanks so much!

Xingyu

ADD REPLYlink written 18 months ago by Xingyu0

We were happy with how the whole approach was working in the different applications shown in the paper. This, however, does not mean that the choice we made is the best one for every possible application and there might be room for improvement by making a more careful choice of bandwidth. If users feel they want to tune that parameter, we can add an argument to do it. If you want to have a deeper understanding of this subject, you should start with the book of Silverman (1986).

cheers,

robert.

ADD REPLYlink written 18 months ago by Robert Castelo2.3k

Thanks very much. I'm now trying to tune the Sigma parameter '4' in the GSVA package. I directly changed this parameter in the src C code of GSVA source package. Hope that will work.

ADD REPLYlink written 18 months ago by Xingyu0
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 16.09
Traffic: 293 users visited in the last hour