I would add that the reason the TopTable results do not agree with the 2-fold or more results, is that generally statistical tests compare treatment mean differences to within treatment variation. Hence, if the results are not variable, you will have many statistically significant genes that have less than 2-fold difference. As mentioned many times on this list, statistical significance does not imply biological significance, but differences that are not statistically significant may be due to chance variation and thus are unlikely to have biological significance. The converse side of this is that if genes are highly variable, they may have more than 2-fold difference and not be statistically significant. The purpose of normalization is to remove biases that differ from array to array due to the hybridization and labeling processes, so that comparisons between conditions are free of this part of the experimental error. This improves our power to detect statistically significant differential expression. --Naomi At 09:59 AM 4/13/2005, Gordon Barr wrote: >Gorjanc and Vijay > >This is a misconception as to why to normalize the data. It is not so >that we can get "pleasing" results or agreement between analytic >methods but because statistically it is the correct thing to do. If I >use the wrong statistical test on a set of data (e.g. parametric tests >on data that violates all the assumptions) and it gives the same >result as an appropriate non-parametric analysis that does not make it >"right" and ok to do again. It means I got lucky. If the analysis of >non-normalized data is the same as of normalized data you are lucky not >right. Sean is on target- if they agree normalize; if they do not agree >normalize. I would add to that why bother analyzing the non- normalized >data. > >Gordon > > >Gordon A. Barr, Ph.D. >Senior Research Scientist >NYS Psychiatric Institute >Columbia College of Physicians and Surgeons >212-543-5694 (V) >212-543-5467 (F) >"There is no flag large enough to cover the shame of killing innocent >people." -- Howard Zinn >_____________________________________________________ >This e-mail is confidential and may be privileged. Use or disclosure >of it by anyone other than a designated addressee is unauthorized. If >you are not an intended recipient, please delete this e-mail. > >On Apr 13, 2005, at 8:29 AM, Gorjanc Gregor wrote: > >>>-----Original Message----- >>>From: Sean Davis [mailto:sdavis2@mail.nih.gov] >>>Sent: sre 2005-04-13 14:16 >>>To: Gorjanc Gregor >>>Cc: bioconductor@stat.math.ethz.ch >>>Subject: Re: [BioC] is-normalisation-really-required >>> >>>On Apr 13, 2005, at 8:03 AM, Gorjanc Gregor wrote: >>> >>>>Hi! >>>> >>>>You might try analysis with and without normalization and take >>>>a look at the results. If they say the same thing than I would >>>>say, no it is not necessary to do normalization. >>> >>>So, if the two results agree, then the results with normalization are >>>correct; if not then the results with normalization are still correct. >>>Sounds like we are pretty much stuck with normalization.... >>> >>>Sean >>Why should one do normalization if the results aren't different. But, >>in >>that case it really does not matter and one can do it or not. >> >> >>>>dear friends >>>>i have situation, where i thought its ok for me not to >>>>do normalisation, i am afraid i may be wrong. i want >>>>your advice in this regard. >>>> >>>>we performed a wild type - mutant, dye-swap >>>>experiment. >>>>when we analysed the intensity values, they were >>>>consistant among the two experiment (dye-swap). ie., >>>>almost same values for mutants in both the experiments >>>>of the dye-swap. >>>>since the values are almost same, i thought there >>>>might not be any dye-bias, so i just went ahead, >>>>averaged the two values, found out their ratio and >>>>filtered genes with 2 fold change. >>>> >>>>so i have done this without normalisation. >>>>i am afraid, i might be wrong, my 2 fold chaging genes >>>>might be wrong... >>>>kindly give me your advice in this regard. >>>>i did analyse the data with limma, but the topTable >>>>genes there never correlates with my 2 fold genes. >>>> >>>>kindly correct me. >>>>thanks >>>> >>>>vijay >>>>graduate student >>>>department of biological sciences >>>>the university of southern mississippi >>>>MS, USA >>> >>>-- >>>Lep pozdrav / With regards, >>> Gregor Gorjanc >>> >>>------------------------------------------------------------------- --- - >>>- >>>University of Ljubljana >>>Biotechnical Faculty URI: http://www.bfro.uni- lj.si/MR/ggorjan >>>Zootechnical Department email: gregor.gorjanc <at> bfro.uni- lj.si >>>Groblje 3 tel: +386 (0)1 72 17 861 >>>SI-1230 Domzale fax: +386 (0)1 72 17 888 >>>Slovenia >>> >>>_______________________________________________ >>>Bioconductor mailing list >>>Bioconductor@stat.math.ethz.ch >>>https://stat.ethz.ch/mailman/listinfo/bioconductor >> >>_______________________________________________ >>Bioconductor mailing list >>Bioconductor@stat.math.ethz.ch >>https://stat.ethz.ch/mailman/listinfo/bioconductor > >_______________________________________________ >Bioconductor mailing list >Bioconductor@stat.math.ethz.ch >https://stat.ethz.ch/mailman/listinfo/bioconductor Naomi S. Altman 814-865-3791 (voice) Associate Professor Bioinformatics Consulting Center Dept. of Statistics 814-863-7114 (fax) Penn State University 814-865-1348 (Statistics) University Park, PA 16802-2111

Naomi Altman wrote: > I would add that the reason the TopTable results do not agree with the > 2-fold or more results, is that generally statistical tests compare > treatment mean differences > to within treatment variation. Hence, if the results are not variable, > you will have many statistically significant genes that have less than > 2-fold difference. As mentioned many times on this list, statistical > significance does not imply biological significance, but differences > that are not statistically significant may be due to chance variation > and thus are unlikely to have biological significance. The converse > side of this is that if genes are highly variable, they may have more > than 2-fold difference and not be statistically significant. > > The purpose of normalization is to remove biases that differ from array > to array due to the hybridization and labeling processes, so that > comparisons between conditions are free of this part of the experimental > error. This improves our power to detect statistically significant > differential expression. > Naomi, thanks for the update! >> Gorjanc and Vijay >> >> This is a misconception as to why to normalize the data. It is not so >> that we can get "pleasing" results or agreement between analytic >> methods but because statistically it is the correct thing to do. If I >> use the wrong statistical test on a set of data (e.g. parametric tests >> on data that violates all the assumptions) and it gives the same >> result as an appropriate non-parametric analysis that does not make it >> "right" and ok to do again. It means I got lucky. If the analysis of >> non-normalized data is the same as of normalized data you are lucky not >> right. Sean is on target- if they agree normalize; if they do not agree >> normalize. I would add to that why bother analyzing the non- normalized >> data. >> >> Gordon >> >> >> Gordon A. Barr, Ph.D. >> Senior Research Scientist >> NYS Psychiatric Institute >> Columbia College of Physicians and Surgeons >> 212-543-5694 (V) >> 212-543-5467 (F) >> "There is no flag large enough to cover the shame of killing innocent >> people." -- Howard Zinn >> _____________________________________________________ >> This e-mail is confidential and may be privileged. Use or disclosure >> of it by anyone other than a designated addressee is unauthorized. If >> you are not an intended recipient, please delete this e-mail. >> >> On Apr 13, 2005, at 8:29 AM, Gorjanc Gregor wrote: >> >>>> -----Original Message----- >>>> From: Sean Davis [mailto:sdavis2@mail.nih.gov] >>>> Sent: sre 2005-04-13 14:16 >>>> To: Gorjanc Gregor >>>> Cc: bioconductor@stat.math.ethz.ch >>>> Subject: Re: [BioC] is-normalisation-really-required >>>> >>>> On Apr 13, 2005, at 8:03 AM, Gorjanc Gregor wrote: >>>> >>>>> Hi! >>>>> >>>>> You might try analysis with and without normalization and take >>>>> a look at the results. If they say the same thing than I would >>>>> say, no it is not necessary to do normalization. >>>> >>>> >>>> So, if the two results agree, then the results with normalization are >>>> correct; if not then the results with normalization are still correct. >>>> Sounds like we are pretty much stuck with normalization.... >>>> >>>> Sean >>> >>> Why should one do normalization if the results aren't different. But, >>> in >>> that case it really does not matter and one can do it or not. >>> >>> >>>>> dear friends >>>>> i have situation, where i thought its ok for me not to >>>>> do normalisation, i am afraid i may be wrong. i want >>>>> your advice in this regard. >>>>> >>>>> we performed a wild type - mutant, dye-swap >>>>> experiment. >>>>> when we analysed the intensity values, they were >>>>> consistant among the two experiment (dye-swap). ie., >>>>> almost same values for mutants in both the experiments >>>>> of the dye-swap. >>>>> since the values are almost same, i thought there >>>>> might not be any dye-bias, so i just went ahead, >>>>> averaged the two values, found out their ratio and >>>>> filtered genes with 2 fold change. >>>>> >>>>> so i have done this without normalisation. >>>>> i am afraid, i might be wrong, my 2 fold chaging genes >>>>> might be wrong... >>>>> kindly give me your advice in this regard. >>>>> i did analyse the data with limma, but the topTable >>>>> genes there never correlates with my 2 fold genes. >>>>> >>>>> kindly correct me. >>>>> thanks >>>>> >>>>> vijay >>>>> graduate student >>>>> department of biological sciences >>>>> the university of southern mississippi >>>>> MS, USA >>>> >>>> >>>> -- >>>> Lep pozdrav / With regards, >>>> Gregor Gorjanc >>>> >>>> ---------------------------------------------------------------------- >>>> - >>>> - >>>> University of Ljubljana >>>> Biotechnical Faculty URI: http://www.bfro.uni- lj.si/MR/ggorjan >>>> Zootechnical Department email: gregor.gorjanc <at> bfro.uni- lj.si >>>> Groblje 3 tel: +386 (0)1 72 17 861 >>>> SI-1230 Domzale fax: +386 (0)1 72 17 888 >>>> Slovenia >>>> >>>> _______________________________________________ >>>> Bioconductor mailing list >>>> Bioconductor@stat.math.ethz.ch >>>> https://stat.ethz.ch/mailman/listinfo/bioconductor >>> >>> >>> _______________________________________________ >>> Bioconductor mailing list >>> Bioconductor@stat.math.ethz.ch >>> https://stat.ethz.ch/mailman/listinfo/bioconductor >> >> >> _______________________________________________ >> Bioconductor mailing list >> Bioconductor@stat.math.ethz.ch >> https://stat.ethz.ch/mailman/listinfo/bioconductor > > > Naomi S. Altman 814-865-3791 (voice) > Associate Professor > Bioinformatics Consulting Center > Dept. of Statistics 814-863-7114 (fax) > Penn State University 814-865-1348 (Statistics) > University Park, PA 16802-2111 > > -- Lep pozdrav / With regards, Gregor Gorjanc ---------------------------------------------------------------------- University of Ljubljana Biotechnical Faculty URI: http://www.bfro.uni-lj.si/MR/ggorjan Zootechnical Department mail: gregor.gorjanc <at> bfro.uni-lj.si Groblje 3 tel: +386 (0)1 72 17 861 SI-1230 Domzale fax: +386 (0)1 72 17 888 Slovenia, Europe ---------------------------------------------------------------------- "One must learn by doing the thing; for though you think you know it, you have no certainty until you try." Sophocles ~ 450 B.C.