Dear Bioc, I want to use the Knn function, from packages(class), to classify sample from a cDNA dataset in which there are some missing values. When I run the script, the function complains about missing values, so I checked for some strategies to overcome the problem. One way could be to fit the missing values on the data itself, with pamr.knnimpute, but I don't like the idea, as I maight overfit the data itself. On the other side I was looking for a strategy to make knn accept missing values but I could not find. do you have any suggestion? -- Claudio ------------------------------------------------------ Leggi GRATIS le tue mail con il telefonino i-mode? di Wind http://i-mode.wind.it/

Dear Claudio, Here is a possible solution. Assign NAs randomly (but with some sense) and classify the samples. Repeat this "NAs-assignment + classification" procedure several times and compare the results. If they are are similar, I would say that there is no risk of overfitting if you assign NAs with knnimpute. If they are not similar, the question remains open. With respect to how to implement all this in R, I cannot help you very much. Best, Federico claudio.is at libero.it wrote: > Dear Bioc, > > I want to use the Knn function, from packages(class), to classify sample from a cDNA dataset in which there are some missing values. > When I run the script, the function complains about missing values, so I checked for some strategies to overcome the problem. > One way could be to fit the missing values on the data itself, with pamr.knnimpute, but I don't like the idea, as I maight overfit the data itself. > On the other side I was looking for a strategy to make knn accept missing values but I could not find. > do you have any suggestion? > > -- > Claudio > > > ------------------------------------------------------ > Leggi GRATIS le tue mail con il telefonino i-mode? di Wind > http://i-mode.wind.it/ > > > _______________________________________________ > 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 > >

dear claudio maybe you have a look at the pcaMethods package. *GENE EXPRESSION:* Wolfram Stacklies, Henning Redestig, Matthias Scholz, Dirk Walther, and Joachim Selbig *pcaMethods?a bioconductor package providing PCA methods for incomplete data* Bioinformatics, 1 May 2007; 23: 1164 - 1167. hth, matthias. claudio.is at libero.it schrieb: >Dear Bioc, > >I want to use the Knn function, from packages(class), to classify sample from a cDNA dataset in which there are some missing values. >When I run the script, the function complains about missing values, so I checked for some strategies to overcome the problem. >One way could be to fit the missing values on the data itself, with pamr.knnimpute, but I don't like the idea, as I maight overfit the data itself. >On the other side I was looking for a strategy to make knn accept missing values but I could not find. >do you have any suggestion? > >-- >Claudio > > >------------------------------------------------------ >Leggi GRATIS le tue mail con il telefonino i-mode? di Wind >http://i-mode.wind.it/ > > >_______________________________________________ >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 > > > -- *************************************************** Dr. Matthias E. Futschik Institute for Theoretical Biology Charite - School of Medicine Humboldt-University Invalidenstra?e 43 10115 Berlin Germany email: m.futschik at staff.hu-berlin.de phone: + 49-30-2093 9106 fax: + 49-30-2093 8801 URL: http://itb.biologie.hu-berlin.de/~futschik