Dear Sean, Thanks for your reply. Probably i should not ask this question across bioc as i dont have any other source to ask, i'm putting this question across to you. I'm sorry for not explaining the problem in my previous email, After seeing your email i figured that i already have TRUE values. We have conducted the following experiment: We have performed DSBR assay and acquired SSF(DNA repair) values and genotype information for Placebo (control) and drug sample. We have a total of 116 samples of which 58 are placebo and 58 are drug samples. The objective of doing ROC is to get an optimal threshold value to find if there is any association between the genotype information and SSF (trait) value for Drug sample and placebo separately. Having that as an objective, using "placebo" and "drug" as classifiers, we will end up with one cut-off value. Using this cut-off I would like to perform chi-sqaure test for drug SSF values and placebo SSF values to get the genotypes associated with more DNA damage (i.e SSF>cut-off will imply more damage) and less damage(i.e SSF<cut-off will="" imply="" dna="" repair).="" if="" that="" is="" the="" case,="" will="" that="" make="" sense="" to="" have="" "placebo"="" and="" "drug"="" as="" classifiers="" in="" roc?="" thanking="" you="" in="" anticipation,="" regards,="" rasanpreet="" kaur="" on="" sat,="" oct="" 30,="" 2010="" at="" 6:26="" pm,="" sean="" davis="" <sdavis2="" at="" mail.nih.gov=""> wrote: > > > On Sat, Oct 30, 2010 at 12:35 AM, rasanpreet kaur suri > <rasanpreet.kaur at="" gmail.com=""> wrote: >> >> Hello friends, >> I want to perform ROC analysis on a list of Strand scission factor values. >> eg: >> SSF_value >> 0.123 >> 0.345 >> 0.235 >> 0.546 >> .. >> >> I have no binary trait to perform the analysis. Could you please let me >> know >> how proper it is to simulate values or classifiers to get a threshold >> value >> and which function i can use to simulate the classifiers. I will be >> thankful >> for any knid of advice. >> > > Hi, Rasanpreet. ?You need to know the "true" values for a classifier so that > you can determine sensitivity and specificity of the test. ?Without those > values, I'm not sure how how ROC curves would be applied. > Sean > >> >> Thanking you in anticipation. >> >> Regards, >> Rasanpreet Kaur >> >> ? ? ? ?[[alternative HTML version deleted]] >> >> _______________________________________________ >> Bioconductor mailing list >> Bioconductor at stat.math.ethz.ch >> https://stat.ethz.ch/mailman/listinfo/bioconductor >> Search the archives: >> http://news.gmane.org/gmane.science.biology.informatics.conductor > > -- Rasanpreet Kaur

On Wed, Nov 3, 2010 at 3:37 AM, rasanpreet kaur suri < rasanpreet.kaur@gmail.com> wrote: > Dear Sean, > > Thanks for your reply. Probably i should not ask this question across > bioc as i dont have any other source to ask, i'm putting this question > across to you. I'm sorry for not explaining the problem in my previous > email, After seeing your email i figured that i already have TRUE > values. > > We have conducted the following experiment: > > We have performed DSBR assay and acquired SSF(DNA repair) values and > genotype information for Placebo (control) and drug sample. We have a > total of 116 samples of which 58 are placebo and 58 are drug samples. > The objective of doing ROC is to get an optimal threshold value to > find if there is any association between the genotype information and > SSF (trait) value for Drug sample and placebo separately. Having that > as an objective, using "placebo" and "drug" as classifiers, we will > end up with one cut-off value. Using this cut-off I would like to > perform chi-sqaure test for drug SSF values and placebo SSF values to > get the genotypes associated with more DNA damage (i.e SSF>cut-off > will imply more damage) and less damage(i.e SSF<cut-off will="" imply="" dna=""> repair). > > If you are looking for genotype differences, then test that directly using a hypothesis test. In this case, it sounds like you have a two-factor setup, treatment and genotype, with genotype being the factor of interest. Why not test for differences in SSF values between the genotypes? A simple place to start is with plotting the data using something like ggplot2 or lattice to allow you to see if there are observable differences. > If that is the case, will that make sense to have "Placebo" and "drug" > as classifiers in ROC? > > I don't think so. I don't see how ROC curves will be useful given what you describe above as your question. I could be mistaken, though. Sean > Thanking you in anticipation, > > Regards, > Rasanpreet Kaur > > On Sat, Oct 30, 2010 at 6:26 PM, Sean Davis <sdavis2@mail.nih.gov> wrote: > > > > > > On Sat, Oct 30, 2010 at 12:35 AM, rasanpreet kaur suri > > <rasanpreet.kaur@gmail.com> wrote: > >> > >> Hello friends, > >> I want to perform ROC analysis on a list of Strand scission factor > values. > >> eg: > >> SSF_value > >> 0.123 > >> 0.345 > >> 0.235 > >> 0.546 > >> .. > >> > >> I have no binary trait to perform the analysis. Could you please let me > >> know > >> how proper it is to simulate values or classifiers to get a threshold > >> value > >> and which function i can use to simulate the classifiers. I will be > >> thankful > >> for any knid of advice. > >> > > > > Hi, Rasanpreet. You need to know the "true" values for a classifier so > that > > you can determine sensitivity and specificity of the test. Without those > > values, I'm not sure how how ROC curves would be applied. > > Sean > > > >> > >> Thanking you in anticipation. > >> > >> Regards, > >> Rasanpreet Kaur > >> > >> [[alternative HTML version deleted]] > >> > >> _______________________________________________ > >> Bioconductor mailing list > >> Bioconductor@stat.math.ethz.ch > >> https://stat.ethz.ch/mailman/listinfo/bioconductor > >> Search the archives: > >> http://news.gmane.org/gmane.science.biology.informatics.conductor > > > > > > > > -- > Rasanpreet Kaur > > _______________________________________________ > Bioconductor mailing list > Bioconductor@stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor > Search the archives: > http://news.gmane.org/gmane.science.biology.informatics.conductor > [[alternative HTML version deleted]]