Thanks, Bill. As a non-statistician but an ardent scientific type , I thoroughly agree with Bill's approach/comments: . Almost everyone uses Excel, so why use proprietary (aka home-grown) software? That just muddies the understanding of results. . Why use just two families? The 111 results must be pouring in, providing rich data for multiple lines of inquiry. . If you have a theory, please posit it in the scientific method and support/not support it with rich and voluminous data. . And most of all: Please try to explain all of the above in non-academic terms. One of Bryan Sykes' gifts has been to do that, and anyone who obfuscates (or does not try to illuminate for the masses) is not serving the greater good. My father's 111 results are almost ready, and I would love to have someone explain-for the sake of newcomers/non-scientists/non-academicians-exactly what the significance is, for example, a 111 -9 match compared to a 67 -3 match. As one who is interested in finding relatives/ancestral lines, these are some of the things that I'm most interested in. I'm not, however, interested in the minutiae of "ONE Conroy and ONE Ewing" and abstruse musings. Thanks, Marie Golden Kerr, on behalf of my father, James J. Golden, a 1st generation Irish-American from Rathlacken, County Mayo, Ireland -----Original Message----- From: dna-r1b1c7-bounces@rootsweb.com [mailto:dna-r1b1c7-bounces@rootsweb.com] On Behalf Of Bill Howard Sent: Sunday, July 10, 2011 3:02 PM To: dna-r1b1c7@rootsweb.com Subject: Re: [R-M222] M222 Tree Sandy, You have obviously not read my FAQs, and I certainly suggest that you do so before doing postings like this one. My criticism still stands. I don't know how you did it. Off line, you said that you would send me an Excel spreadsheet if I could not open yours. I have told you that I could not open the file you sent, and you have not done it. I realize that you don't work in Excel, but it should be easy to make a spreadsheet that can be converted to Excel. Most people can do this since Excel is part of a Microsoft family of applications that are perhaps the widest used around the world. I completely agree that if you are comparing only one Ewing and one Conroy haplotype, you can easily get a difference like the one you cite. But you jump to an unwarranted conclusion about it. A careful reading of my FAQs would indicate that the errors on individual pairs of values can be quite large. What you point out is that the RCCs of two individual test results differ by 21 in RCC when their 37 and 67 markers are compared. That is certainly NOT the same as saying that the overall RCC time scale is different. To me it is not at all disturbing. It merely shows that: . You have based a conclusion on only one anecdotal piece of evidence. A statistician would never do that. . You are comparing only one 37 marker result with only one 67 marker result. A statistician would never do that. . When you correlate 37 markers separately from 67 markers, the results are bound to differ, since the input is different. But, my contention is that there is NO evidence that, over a large set of examples, they will differ significantly. . You are ignoring what I have already said about the 37 and 67 marker trees; they will differ in detail. But not in their general form. . And, of course one anecdotal change between ONE Conroy and ONE Ewing will be more dramatic. In short, a difference of 6 o 7 is in the noise. As I wrote before, you should compare all the haplotypes, not just the Ewings with the rest. You are not doing the analysis as you should be doing. I will make the same suggestions to you that I have done before: . , do it right, . do it so that I can check it, . pay attention to statistics, . compute the SDs of your statements, and, . read my FAQs which are located at: < <http://mysite.verizon.net/weh8/FAQ.pdf> http://mysite.verizon.net/weh8/FAQ.pdf> It doesn't matter how many programming languages you know how to use; it does matter that it is done: . thoroughly, . correctly, and, . in a way that is reproduceable, and, . it can be explained. You have not done that. Until that is done, and until your work is checked and is found to be reproduceable, as all scientific studies must be, I cannot agree with what you have done or the way that you did it. I just want to add that, when we first began an off-line correspondence I suggested that you send me your data and explain your approach, and you backed off. I think you are continuing to do so. - Bye from Bill Howard On Jul 10, 2011, at 2:17 PM, Sandy Paterson wrote: > I have explained to bill (sic!) Bill I think 3 times, that I don't > work in Excel and seldom use spreadsheets. Mostly, I write my own > software in one of 3 programming languages, depending on the > application, and I use files, not spreadsheets, so there is no spreadsheet to send him. > > He doesn't seem to understand that. > > Still, I think what we are discussing is of sufficient importance to > warrant some kind of resolution, so what I'll do is to give a specific > example of what I'm getting at. The example I've chosen is a > comparison between Ewing > 26605 and Conroy 16646, over 37 markers and over 67 markers. > > Correlation coef. over 37 markers 0.990382002396377 > Correlation coef. Over 67 markers 0.992411264610795 > > RCC over 37 markers = (1/.990382002396377 - 1)*10000 = 97.11 RCC > over 67 markers = (1/.992411264610795 - 1)*10000 = 76.47 > > I did this comparing all 19 Ewings in my files to all non-Ewings in my > files and found what I believe is a disturbing result. The mean RCC > falls by 6.71 in changing from 37 to 67 markers. Not a single RCC > increases; they all decrease. > > Clearly the change in RCC between Conroy and Ewing is far more dramatic. > > > Anyone who wishes can check the above by doing the following : > > 1.Place the Ewing 26605 marker values in column A in an Excel spreadsheet. > 2.Place the Conroy 16646 marker values in column B in the same spreadsheet. > 3.Use the CORREL function in Excel, denoting the arrays as A1:A37 and > B1:B37 (changing the 37 to 67 if appropriate). > > The RCC's are calculated by taking the reciprocal of the correlation > coefficient, subtracting 1 then multiplying by 10000, as illustrated above. > > > Sandy > > > > > > -----Original Message----- > From: dna-r1b1c7-bounces@rootsweb.com > [mailto:dna-r1b1c7-bounces@rootsweb.com] On Behalf Of Bill Howard > Sent: 10 July 2011 15:54 > To: dna-r1b1c7@rootsweb.com > Subject: Re: [R-M222] M222 Tree > > Sandy, > > I have suggested a number of times that you should send me your > spreadsheet where you have the details of comparisons like this one > and you have not done so. > > > R1b1c7 Research and Links: > > <http://clanmaclochlainn.com/R1b1c7/> http://clanmaclochlainn.com/R1b1c7/ > ------------------------------- > To unsubscribe from the list, please send an email to > <mailto:DNA-R1B1C7-request@rootsweb.com> DNA-R1B1C7-request@rootsweb.com with the word 'unsubscribe' without > the quotes in the subject and the body of the message R1b1c7 Research and Links: <http://clanmaclochlainn.com/R1b1c7/> http://clanmaclochlainn.com/R1b1c7/ ------------------------------- To unsubscribe from the list, please send an email to <mailto:DNA-R1B1C7-request@rootsweb.com> DNA-R1B1C7-request@rootsweb.com with the word 'unsubscribe' without the quotes in the subject and the body of the message