Никита Николаевич Хромов-Борисов Кафедра физики, математики и информатики

  • View
    48

  • Download
    0

Embed Size (px)

DESCRIPTION

24- - ( ) 5-6 2014 . -. ( / ) , . - , - PowerPoint PPT Presentation

Transcript

/ ,

24- - () 5-6 2014 . -

1

(/ ) , -

, . . .. 8-952-204-89-49 (2); 8-921-449-29-05 ()

Nikita.KhromovBorisov@gmail.comhttp://independent.academia.edu/NikitaKhromovBorisov2 : ()

3

4 5 : (Jenks S., Nancy Volkers N. Razors and refrigerators and reindeer Oh My! //J. Natl. Cancer Inst., 1992. Vol. 84. No. 24. P.1863)

- , , !

6 , , :Buchanan A.V., Weiss K.M., Fullerton S.M. Dissecting complex disease: the quest for the Philosophers Stone? International Journal of Epidemiology 2006. Vol. 35. P. 562571.

7 () - ( ) ,

, ! 8 AB0 AB0. . ; 0 ; 0 , B ; B ; AB0 : ; 0 , A ; 2 IQ; . .9 AB0 , , , , . , () , () (OR) OR=1,5. 10 ( ) OROR 1 1,5 1,5 3,5 3,5 9,09,0 3232 360> 360 11: Hopkins W.G. A Scale of Magnitudes for Effect Statistics http://www.sportsci.org/resource/stats/ .., - .. // , 2013. .11. . 7790. , OR5,4, , (pM>0,3). , ORL. (RR5).IoannidisJ.A.P. Commentary: Grading the credibility of molecular evidence for complex diseases //International Journal of Epidemiology, 2006. Vol. 35. P. 572577.

12Begley C.G., Ellis L.M. Raise standards for preclinical cancer research // Nature, 2012. Vol. 483. P. 531-533. (C. Glenn Begley), - Amgen, (Lee M. Ellis) , () 47 53 , .13 , 1900 , . , .

14

Ioannidis J.P.A. Why most published research findings are false // PLoS Med., 2005. Vol. 2. No. 8. Paper: e124.

?

15 (Karl Raimund Popper;28.07.190217.09.1994) . XX

16 (). : , , .. .17 . , , , , .1819 - , , , , . . , , . , , ( , - , ).

1919Albert Abraham Michelson (19.12.1852 09.05.1931)

Edward Williams Morley (29.01.183924.02.1923)

20 (Gregor Johann Mendel;20.07.1822 06.01.1884) 1884

- . . 2011 .

21 PvalBF0195%- Mendel602220010,9082,00,740,750,76Correns26318670,7752,40,740,740,77Tschermak658021490,4161,70,740,750,76Hurst13104450,7336,20,730,750,77Bateson1190339030,3778,10,7460,7530,760Lock692023720,2444,40,7360,7450,753White 16475430,8242,20,730,750,77Darbishire115811383960,36238,10,7490,7510,7532223 . , , , . , , . .2323 / . : () (). : - , .

2424

25 P-26 , Pval, () .2728 P P , : tobs = |tobs| T* , 0, , H0: Pval = Pr(T* tobs|H0). Pval .

28 Pval , , , , . , . , 0 1.

29n1 = 5, n2 = 7, df = 10, t = 1,5P = 0,16 30

n1 = 5, n2 = 7, df = 10, t = 3,0 P = 0,013 = 0,05, 0,0131

() Pval 0,050,05 0,01 *0,01 0,001**0,001 0,0001 ***< 0,0001 ****4- () 0,0001 : http://www.graphpad.com/guides/prism/6/statistics/index.htm?interpreting_a_small_p_value_from_an_unpaired_t_test.htm 3839[0,05; 0,01] .., .., .., - .. , .: - , 1982. 264 . , . :) Pval > 0,05, ;) Pval < 0,01, ;) 0,01 < Pval < 0,05, .39

39 () Pval, () () .

4040 Pval, , , , , , , -.4141 Pval, , , , .

424243 P- , ( ). , , . .434344 () : P-, H0, ; H0. ( ) P- . 444445 P- ! P- , H0:Pval = Pr{T t.|H0}, :P{t|H0} P{H0|t}45P(L|D) P(D|L) , , - (D), (L) , 100%:P(L|D) = 100% , , : P(D|L) = 10-6 = 0,0001%46 Pval , Pval, . :http://en.wikipedia.org/wiki/P-value . , Pval:Goodman S. A dirty dozen: Twelve P-value misconceptions // Semin. Hematol., 2008. - Vol. 45. P. 135-140.

4748 () Pval 484849 () : H0: 1 = 2 , :0 = 1 - 2 = 0 = 1 - 2., 100(1 )%- 0 = 0, unkn 0 = 0 , . , unkn ., unkn ., unkn . 100(1 )%- unkn:5051 5151 , , . (International Committee of Medical Journal Editors ICMJE). , 2013 .: http://www.icmje.org/index.html 2005 .:http://www.mediasphera.ru/mjmp/2005/5/10.pdf

52 , , , , . , (, ). , , P, . , .53Francis Galton, 1901"I have begun to think that no one ought to publish biometric results, without lodging a well-arranged and well-bound copy of his data in some place where it should be accessible, under reasonable restrictions, to those who desire to verify his work. , , ( ) , .

54 - 54 40 ( .. ) 55

dC , Pval , BF01 - ( /) H1 H0, , . 56 P* Pval P- ()

. H0 , Pval [0; 1]. :

565657 , , , . , .5757 Pval Cumming, G. (2008). Replication and p intervals: p values predict the future only vaguely, but confidence intervals do much better // Perspectives on Psychological Science, 2008. Vol. 3. No. 4. P. 286-300. ESCI PPS p intervals http://www.latrobe.edu.au/psy/esci/ 12 25 diff = 0 = 0,05, diff diff = 10 585880%- PvalCumming G. Replication and p intervals: p values predict the future only vaguely, but condence intervals do much better // Perspectives on Psychological Science, 2008. Vol. 3. No. 4. P. 186-300. Pobs80%- Pval 0,05 810-5 0,440,01 610-6 0,220,001 210-7 0,075960 P P 80%- Pval P(H0) repr0,050,44> 30%< 50%0,010,22> 10%< 73%0,0010,07> 2%< 90% . Sellke T., Bayarri M.J., Berger J.O. Calibration of p values for testing precise null hypotheses. The American Statistician, Vol. 55, No. 1. (2001), pp. 62-71.Goodman S.N. A comment on replication, p-values and evidence // Statistics in Medicine, 1992. Vol. 11. P. 875-879.6060 Pval , Pval. Pval , , , , , .. 61 = 0,05 ( = 0,01 ) (Pval < 0,05) ( ) . = 0,001, .. Pval < 0,001, .

62- P < 0,05 ( 0,001 < P < 0,05).- .- , .

6363 P-

6464 .

Lambdin C. Significance tests as sorcery: Science is empiricalsignificance tests are not. Theory & Psychology, 2912. 22(1): 67 90.65Campbell J. P. Editorial: Some remarks from the outgoing editor. Journal of Applied Psychology, 1982. 67: 691-700. . , , , , . 66 67 . , , . .

68 , Pval < 0,05 .

69 (John Wilder Tukey, 16.04.1915 26.07.2000) . () . ( ) . .70

7071 717172 (, , , , , . .) , (, ) . , , . , , .7273 () () , , .. , .737374 (Cohen) dC

7474 , dC

7576 dC http://www.sportsci.org/resource/stats/ , dC 0 0,20,2 0,50,5 1,01,0 2,02,0 4,0 4,0 - 7676 : 7778 , BF BF val. , , :BF01 = P(Dobs|H0) / P(Dobs|H1)BF10 = P(Dobs|H1) / P(Dobs|H0) , , . .

, BF10 BF01BF01 0 1>10030 100 10 303 10 ()1 3 BF10 1 079 () - John Arbuthnot29.04.1667 27.02.1735

8080 82 1629 5218 > 4683 1650 2890 > 27221630 4858 > 4457 3231 > 2840 4422 > 4102 3220 > 2908 4994 > 4590 3196 > 2959 5158 > 4839 3441 > 3179 5035 > 4820 3655 > 3349 5106 > 4928 3668 > 3382 4917 > 4605 3396 > 3289 4703 > 4457 3157 > 3013 5359 > 4952 3209 > 2781 5366 > 4784 1660 3724 > 32471640 5518 > 5332 4748 > 4107 5470 > 5200 5216 > 4803 5460 > 4910 5411 > 4881 4793 > 4617 6041 > 5881 4107 > 3997 5114 > 4858 4047 > 3919 4678 > 4319 3768 > 3395 5616 > 5322 3796 > 3536 6073 > 5560 3363 > 3181 1669 6506 > 58291649 3079 > 2746 1670 6278 > 5719 1691 7662 > 7392 6449 > 6061 7602 > 7316 6443 > 6120 7676 > 7483 6073 > 5822 6985 > 6647 6113 > 5738 7263 > 6713 6058 > 5717 7632 > 7229 6552 > 5847 8062 > 7767 6423 > 6203 8426 > 7626 6568 > 6033 7911 > 7452 6247 > 6041 1700 7578 > 70611680 6548 > 6299 8102 > 7514 6822 > 6533 8031 > 7656 6909 > 6744 7765 > 7683 7577 > 7158 6113 > 5738 7575 > 7127 8366 > 7779 7484 > 7246 7952 > 7417 7575 > 7119 8379 > 7687 7737 > 7214 8239 > 7623 7487 > 7101 7840 > 7380 7604 > 7167 1710 7640 > 72881690 7909 > 7302 484 382 > 454 041 938 423 8181 () :Pval 10-8BF01 = 810117C . : 810117 (H0) , (H1) . : () 810117 ().8283 8383

84 . (.. .). Instat+ http://www.reading.ac.uk/ssc/n/n_instat.htm

85n1 = 390; n2 = 300, t = 3,56; Pval = 0,00083; ES = 0,351,532,72; dC = 0,110,260,40 Bayes Factor Calculator http://pcl.missouri.edu/bayesfactor

BF01 = 0,065 BF10 = 1/BF01 = 15,3 15 , H1, , H0 H1 H0