量的データを扱うときに知っておきたいこと —データの分布、有意性、効果量、信頼区間—

Slide 1

Slide 1 text

ྔతσʔλΛѻ͏ͱ͖ʹ ஌͓͖͍ͬͯͨ͜ͱ —σʔλͷ෼෍ɺ༗ҙੑɺޮՌྔɺ৴པ۠ؒ— Ӝ໺ ݚʢ๺ւֶԂେֶʣ email: [email protected] ֎ࠃޠڭҭϝσΟΞֶձୈ56ճશࠃݚڀେձ ˏૣҴాେֶ ɹɹ2016. 8. 7. ຊ೔ͷࢿྉ http://bit.ly/let2016ws

Slide 2

Slide 2 text

Έͳ͞Μʹ࣭໰

Slide 3

Slide 3 text

• ࿦จ౳ͰޮՌྔΛใࠂͨ͜͠ͱ͸͋Γ·͔͢ • ৴པ۠ؒΛใࠂͨ͜͠ͱ͸͋Γ·͔͢ • ޮՌྔ d ΍ η2 ͷܭࢉํ๏Λ஌͍ͬͯ·͔͢ • ඪ४ภࠩͷܭࢉํ๏Λ஌͍ͬͯ·͔͢ Έͳ͞Μʹ࣭໰

Slide 4

Slide 4 text

ՍۭͷσʔλΛ ༻ҙ͠·ͨ͠

Slide 5

Slide 5 text

Name* Test A খ૔ ರ 70 Տ੢ େޒ࿠ 38 খਿ Ꮺ 58 ௶Ҫ ج༞ 48 ӬҪ ج༞ 28 ڮޱ ๏࢚ 54 ݪ ཽ໵ 58 ޿੉ ༎ 38 ౻ా ࢰಐ 42 ຊؒ խ඙ 47 ٶ࡚ ৎ༤ 78 ଜҪ ࿣࿠ 68 ࢁ࡚ ޹ଠ࿠ 40 ԣҪ ޛࢤ 50 ґా ༸հ 68 एࢁ ప 57 ༗അ Ղ೫ 64 ࿨ઘ ګࢠ 76 ؠҪ ඒՂ 43 ߐ઒ ༝Ӊ 90 ਆ୩ ࣿق 58 ๺઒ ઍՂࢠ 38 ࡔా Ѫࡊ 38 ਿా ඒՂ 43 ⁋ຊ ᜫ 58 ౔୩ ே߳ 60 Ӭ੉ ͘ΔΈ 48 দ໦ ಹಸ 45 ଜҪ ݁ࢠ 24 ए௬ ·Έ 36 *ʮͳΜͪΌͬͯݸਓ৘ใʯͰੜ੒ http://kazina.com/dummy/

Slide 6

Slide 6 text

Group A Test A খ૔ ರ 70 Տ੢ େޒ࿠ 38 খਿ Ꮺ 58 ௶Ҫ ج༞ 48 ӬҪ ج༞ 28 ڮޱ ๏࢚ 54 ݪ ཽ໵ 58 ޿੉ ༎ 38 ౻ా ࢰಐ 42 ຊؒ խ඙ 47 ٶ࡚ ৎ༤ 78 ଜҪ ࿣࿠ 68 ࢁ࡚ ޹ଠ࿠ 40 ԣҪ ޛࢤ 50 ґా ༸հ 68 एࢁ ప 57 ༗അ Ղ೫ 64 ࿨ઘ ګࢠ 76 ؠҪ ඒՂ 43 ߐ઒ ༝Ӊ 90 ਆ୩ ࣿق 58 ๺઒ ઍՂࢠ 38 ࡔా Ѫࡊ 38 ਿా ඒՂ 43 ⁋ຊ ᜫ 58 ౔୩ ே߳ 60 Ӭ੉ ͘ΔΈ 48 দ໦ ಹಸ 45 ଜҪ ݁ࢠ 24 ए௬ ·Έ 36 Group B Test A ؠӬ ࿡࿠ 52 ২໦ ҭೋ 59 ย੉ ཽ໵ 61 ࡔݩ ᠳଠ 76 ౡଜ ༏ 45 ৓ా ௕ར 68 ௕୔ ஌࢙ 63 দҪ Ұಙ 69 ࡾݪ ༟࣍࿠ 43 क԰ ཽ࣍ 51 ੨໺ Έ͋ 36 ஑୩ ༏ 51 ؠ୩ ౧ࢠ 39 ্ݪ ܠࢠ 71 ߐޱ Ί͙Έ 26 ٴ઒ ͳͭΈ 79 େ௩ ·͞Έ 55 Ԭ໺ ࿏ࢠ 61 ֯ా ౧ࢠ 89 ઒୺ ݁ҥ 51 ਆށ ࡊʑඒ 71 ֎ࢁ Έ͋ 63 রҪ Έ͖ 41 ࠜ؛ ༏ 41 ࠜ؛ ྱࢠ 83 Ӌా ѥر 93 ෱࢜ ΈΏ͖ 47 ෍ࢪ ༑߳ 37 ଜా จੈ 52 ٢Ӭ ܙས߳ 41

Slide 7

Slide 7 text

Slide 8

Slide 8 text

Slide 9

Slide 9 text

ൺ΂ͯΈΑ͏ σʔλͷཁ໿

Slide 10

Slide 10 text

ൺ΂ͯΈΑ͏ Group A Group B ฏۉ஋ 52.1 57.1 ඪ४ภࠩ 15.1 16.4

Slide 11

Slide 11 text

ൺ΂ͯΈΑ͏ Group A Group B ฏۉ஋ 52.1 57.1 ඪ४ภࠩ 15.1 16.4 ෼෍ͷ޿͕Γ σʔλͷத৺

Slide 12

Slide 12 text

ൺ΂ͯΈΑ͏ Group A Group B 0 20 40 60 80 100 0.000 0.005 0.010 0.015 0.020 0.025 0.030 0 20 40 60 80 100 0.000 0.005 0.010 0.015 0.020 0.025 0.030

Slide 13

Slide 13 text

ൺ΂ͯΈΑ͏ Group A Group B 0 20 40 60 80 100 0.000 0.005 0.010 0.015 0.020 0.025 0.030 0 20 40 60 80 100 0.000 0.005 0.010 0.015 0.020 0.025 0.030

Slide 14

Slide 14 text

ൺ΂ͯΈΑ͏ ฏۉ஋ͷࠩ Group A Group B Group A Group B ฏۉ஋ 52.1 57.1 ඪ४ภࠩ 15.1 16.4 0 20 40 60 80 100 0.000 0.005 0.010 0.015 0.020 0.025 0.030 0 20 40 60 80 100 0.000 0.005 0.010 0.015 0.020 0.025 0.030 ෼෍ͷҧ͍

Slide 15

Slide 15 text

ඪ४ภࠩ

Slide 16

Slide 16 text

ඪ४ภࠩ ෼෍ͷ޿͕Γʢ͹Β͖ͭʣ

Slide 17

Slide 17 text

• ݸʑͷ਺஋ͱฏۉ஋ͱͷࠩΛ̎৐͠ɺ  ͦͷ߹ܭΛσʔλͷ਺Ͱׂͬͨ΋ͷͷฏํࠜ ඪ४ภࠩ Group A Test A খ૔ ರ 70 Տ੢ େޒ࿠ 38 খਿ Ꮺ 58 ௶Ҫ ج༞ 48 ӬҪ ج༞ 28 ڮޱ ๏࢚ 54 ݪ ཽ໵ 58 ޿੉ ༎ 38 ౻ా ࢰಐ 42 (70-52.1)2 = 320.4 (38-52.1)2 = 198.8 (58-52.1)2 = 034.8 . . . ߹ܭ 6828.7 /30 = 227.6 √ 227.6 = 15.1 Group A ฏۉ஋ 52.1 ඪ४ภࠩ 15.1 ←ʢ෼ࢄʣ

Slide 18

Slide 18 text

0 20 40 60 80 100 0.00 0.01 0.02 0.03 0.04 0 20 40 60 80 100 0.00 0.01 0.02 0.03 0.04 ฏۉ஋ = 50 ͷ৔߹ ඪ४ภࠩ = 10 ඪ४ภࠩ = 20 34.1% 13.6% 34.1% 34.1% 13.6% 34.1% 13.6% 13.6% ඪ४ภࠩ

Slide 19

Slide 19 text

0 20 40 60 80 100 0.00 0.01 0.02 0.03 0.04 0 20 40 60 80 100 0.00 0.01 0.02 0.03 0.04 ฏۉ஋ = 50 ͷ৔߹ ඪ४ภࠩ ඪ४ภࠩ = 10 ඪ४ภࠩ = 20

Slide 20

Slide 20 text

ʢ٢ా, 1998, p. 173ʣ ඪ४ภࠩ

Slide 21

Slide 21 text

ʢ٢ా, 1998, p. 173ʣ ඪ४ภࠩ ࠩ͸ಉ͡

Slide 22

Slide 22 text

ඪ४ภࠩ ॏͳΓͷྔ͕ҧ͏

Slide 23

Slide 23 text

ݴ͑ͦ͏ͳ͜ͱ • ฏۉ஋ͷൺֱ͚ͩͰ͸ෆे෼ • σʔλͷ෼෍ʢ޿͕Γʣ΋߹Θͤͯݕ౼ • ෼෍ͷॏͳΓ͕গͳ͍ํ͕͕ࠩେ͖͍

Slide 24

Slide 24 text

΋͏Ұ౓ൺ΂ͯΈΑ͏ Group A Group B 0 20 40 60 80 100 0.000 0.005 0.010 0.015 0.020 0.025 0.030 0 20 40 60 80 100 0.000 0.005 0.010 0.015 0.020 0.025 0.030

Slide 25

Slide 25 text

΋͏Ұ౓ൺ΂ͯΈΑ͏ 0" 1" 2" 3" 4" 5" 6" 7" 8" 9" 0,10" 11,20" 21,30" 31,40" 41,50" 51,60" 61,70" 71,80" 81,90" 91,100" Group"A" Group"B" ࣮ࡍͷ෼෍Λϓϩοτͨ͠΋ͷ

Slide 26

Slide 26 text

΋͏Ұ౓ൺ΂ͯΈΑ͏ 0" 1" 2" 3" 4" 5" 6" 7" 8" 9" 0,10" 11,20" 21,30" 31,40" 41,50" 51,60" 61,70" 71,80" 81,90" 91,100" Group"A" Group"B" ͜ͷॏͳΓ͸େ͖͍ͷʁখ͍͞ͷʁ

Slide 27

Slide 27 text

ࢦඪ͕΄͍͠

Slide 28

Slide 28 text

ޮՌྔʢEﬀect Sizeʣ

Slide 29

Slide 29 text

ޮՌྔʢEﬀect Sizeʣ • ޮՌͷେ͖͞Λ͋ΒΘ͢౷ܭతͳࢦඪ  ʢେٱอɾԬా, 2012, p. 44ʣ • ඪຊαΠζʢe.g., ਓ਺ʣʹґଘ͠ͳ͍

Slide 30

Slide 30 text

༗ҙੑͱޮՌྔ

Slide 31

Slide 31 text

Figure 1. (N = 10) Figure 2. (N = 119) r = .627 r = .184 p = .052 p = .045 > ༗ҙͰͳ͍ʛ༗ҙ 0 2 4 6 8 10 0 2 4 6 8 10 0 3 6 9 12 15 0 3 6 9 12 15

Slide 32

Slide 32 text

ʮʢ౷ܭతʣ༗ҙʯͱ͸ • ໨ͷલͷσʔλʢඪຊʣ͔ΒΑΓେ͖ͳจ຺ ʢ฼ूஂʣΛਪఆ͢Δ • ඪຊͰ؍࡯͞ΕΔࠩɾؔ܎͕ɺ฼ूஂ͔Βͷ ඪຊநग़࣌ͷޡࠩͰੜ͡Δ֬཰ʢp ஋ʣΛ  ܭࢉ͢Δ • p ஋͕ج४஋ʢྟք஋ʣҎԼͰ͋Ε͹ʮ༗ҙʯ Ͱ͋ΔʢޡࠩͰ͸ͳ͍ʣͱ൑அ͢Δ

Slide 33

Slide 33 text

ʮʢ౷ܭతʣ༗ҙʯͱ͸ • ૬ؔ෼ੳͷ৔߹ɺඪຊͰݟΒΕΔ૬͕ؔແ૬ ؔʢ૬ؔθϩʣͷ฼ूஂ͔Βͷඪຊநग़࣌ͷ ޡࠩͰੜ͡Δ֬཰ʢp ஋ʣΛܭࢉ͢Δ

Slide 34

Slide 34 text

฼ूஂ ඪɹຊ ਪఆ σʔλղੳ Σ, F, t, p... ฼ूஂͱඪຊ

Slide 35

Slide 35 text

• ͋ΔඪຊͰಘΒΕͨ୅ද஋ʢe.g., ฏۉʣ ͱ฼ूஂͷ୅ද஋ͱͷࠩ ඪຊޡࠩ

Slide 36

Slide 36 text

฼ूஂ  μ = 15.3 ඪຊA  M = 14.7 ඪຊB  M = 15.9 ඪຊC  M = 15.2 ඪຊD  M = 15.4 ඪຊE  M = 15.1

Slide 37

Slide 37 text

฼ूஂ  μ = 14.7 ඪຊA  M = 14.7 ࣮ࡍ͸ M = μ ͱͯ͠ਪఆ

Slide 38

Slide 38 text

• ඪຊͷαΠζ͕େ͖͚Ε͹େ͖͍΄Ͳɺ ඪຊޡࠩ͸খ͘͞ͳΔ • ͭ·Γਪఆͷਫ਼౓͕ߴ͘ͳΔ ඪຊޡࠩ

Slide 39

Slide 39 text

N 3 4 5 10 20 30 ྟք஋ (α = .05) 0.997 0.950 0.878 0.632 0.444 0.361 N 40 50 100 200 500 1000 ྟք஋ (α = .05) 0.312 0.279 0.197 0.139 0.088 0.062 *͜ΕΑΓେ͖͍਺஋ͩͬͨΒۮવͰͳ͍ͱΈͳ͢ ૬ؔ܎਺ʢrʣͱྟք஋*

Slide 40

Slide 40 text

ʢ٢ా, 1998, p. 75ʣ ૬ؔ܎਺ʢrʣͷҙຯ

Slide 41

Slide 41 text

Figure 1. (N = 10) Figure 2. (N = 119) r = .627 r = .184 p = .052 p = .045 0 2 4 6 8 10 0 2 4 6 8 10 0 3 6 9 12 15 0 3 6 9 12 15 ΋͏Ұ౓ൺ΂ͯΈΑ͏

Slide 42

Slide 42 text

༗ҙͳࠩʢ·ͨ͸ɺ༗ҙͳ૬ؔʣ͕ಘΒΕ΍͍͢ݚڀΛ͢ΔͨΊ ͷ͖ΘΊͯ༗ޮͳํ๏͕͋Γ·͢ɻͦΕ͸ɺͱʹ͔͘ଟ͘ͷσʔ λΛूΊΔ͜ͱͰ͢ɻͳͥͳΒ͹ɽɽɽ౷ܭతݕఆͷ݁Ռ͸σʔ λ਺͕ଟ͍΄Ͳ༗ҙʹͳΓ΍͍͔͢ΒͰ͢ɻͦͷͨΊɺσʔλ਺ Λ૿΍͑͢͠͞Ε͹ɺ͖ΘΊͯখ͞ͳࠩͰ΋ “ʢ౷ܭతʹ͸ʣ༗ ҙͰ͋Δ” ͱ͍͏݁ՌʹͳΔՄೳੑ͕ߴ·Γ·͢ɻྫ͑͹ɺN = 1000 ͷ৔߹ʹ͸ɺr = 0.062 ͱ͍͏͖ΘΊͯখ͞ͳ૬ؔ܎਺ʢ͢ ͳΘͪɺඇৗʹऑ͍ؔ܎ʣͰ΋༗ҙʹͳΓ·͢ɽɽɽɻ͜ͷΑ͏ ʹɺ౷ܭతਪఆʹ͸ɺ“σʔλ਺ͱ͍͏ɺݚڀऀ͕೚ҙʹܾΊΒ ΕΔཁҼʹΑͬͯ݁Ռ͕ࠨӈ͞Εͯ͠·͏” ͱ͍͏ࠜຊతͳ໰୊ ͕͋Γ·͢ɻʢ٢ా, 1998, p. 232; Լઢ͸Ӝ໺ʹΑΔʣ ༗ҙੑݕఆͱඪຊαΠζ

Slide 43

Slide 43 text

Ͳ͏͢Δʁ

Slide 44

Slide 44 text

• ࠩ΍ؔ܎Λද͢ɺඪຊαΠζʹ  ґଘ͠ͳ͍ࢦඪ͕΄͍͠ • ͦΕ͕ޮՌྔʢeﬀect sizeʣ • ࣮͸૬ؔ܎਺ r ΋ޮՌྔͷࢦඪʢޙड़ʣ

Slide 45

Slide 45 text

ޮՌྔͷछྨ

Slide 46

Slide 46 text

• ࠩͷେ͖͞Λද͢ࢦඪʢd ଒ʣ • ؔ܎ͷڧ͞Λද͢ࢦඪʢr ଒ʣ େ͖͘෼͚ͯ̎ͭ

Slide 47

Slide 47 text

ࠩͷେ͖͞Λද͢ࢦඪ Cohen’s d

Slide 48

Slide 48 text

pooled SD X X d 2 1 − = ←ɹฏۉͷࠩ ←ɹඪ४ภࠩ Cohen’s d ʮ̎ͭͷάϧʔϓͷࠩ͸ඪ४ภࠩԿݸ෼ʯ

Slide 49

Slide 49 text

pooled SD X X d 2 1 − = | 52.1 - 57.1| = (15.1 + 16.4) / 2* *ඪຊαΠζ͕ҟͳΔͱ͖ɺSDpooled ͷܭࢉ͸΋͏গ͠ෳࡶʹͳΓ·͢ Cohen’s d Group A Group B ฏۉ஋ 52.1 57.1 ඪ४ภࠩ 15.1 16.4

Slide 50

Slide 50 text

pooled SD X X d 2 1 − = 5.0 = 15.75 *ඪຊαΠζ͕ҟͳΔͱ͖ɺSDpooled ͷܭࢉ͸΋͏গ͠ෳࡶʹͳΓ·͢ Cohen’s d Group A Group B ฏۉ஋ 52.1 57.1 ඪ४ภࠩ 15.1 16.4 = 0.32

Slide 51

Slide 51 text

d 0 0.1 0.2 0.3 0.4 0.5 0.6 ॏͳΓ ʢˋʣ 100 92.3 85.7 78.7 72.6 67 61.8 d 0.7 0.8 0.9 1 1.1 1.2 1.3 ॏͳΓ ʢˋʣ 57 52.6 48.4 44.6 41.1 37.8 34.7 ޮՌྔ d ͱ෼෍ͷॏͳΓ

Slide 52

Slide 52 text

0" 1" 2" 3" 4" 5" 6" 7" 8" 9" 0,10" 11,20" 21,30" 31,40" 41,50" 51,60" 61,70" 71,80" 81,90" 91,100" Group"A" Group"B" d = 0.32 ͳͷͰॏͳΓ͸ 3/4 ͙Β͍ ࠶ͼ͜ͷάϥϑ

Slide 53

Slide 53 text

ͭ·Γ

Slide 54

Slide 54 text

Group A ͱ Group B ͸ɺ ฏۉ఺ʹ 5 ఺͕ࠩ͋Δ͕ɺ શମͷ 3/4 ͸ॏͳ͍ͬͯΔɻ

Slide 55

Slide 55 text

ޮՌྔ d ͱॏͳΓͷؔ܎

Slide 56

Slide 56 text

pooled SD X X d 2 1 − = ←ɹখ͍͞ํ͕ྑ͍ ←ɹେ͖͍ํ͕ྑ͍ d ஋͕େ͖͘ͳΔʹ͸ ʮฏۉͷ͕ࠩେ͖͘ɺඪ४ภ͕ࠩখ͘͞ͳΔͱɺ ޮՌྔ͸େ͖͘ͳΔɻʯ

Slide 57

Slide 57 text

ޮՌྔͷղऍ

Slide 58

Slide 58 text

• Cohen (1988) • small: d = 0.2, overlap: 85.7% • e.g., 15ࡀͱ16ࡀͷঁࢠͷ਎௕ࠩ • medium: d = 0.5, overlap: 67.0% • e.g., 14ࡀͱ18ࡀͷঁࢠͷ਎௕ࠩ • large: d = 0.8, overlap: 52.6% • e.g., େֶ৽ೖੜͱPhDऔಘऀͷIQࠩ ޮՌྔͷղऍ

Slide 59

Slide 59 text

• Plonsky & Oswald (2014) • “L2 field-specific benchmarks” ޮՌྔͷղऍ ܈ؒൺֱ ܈಺ൺֱ small d = 0.40 d = 0.60 medium d = 0.70 d = 1.00 large d = 1.00 d = 1.40

Slide 60

Slide 60 text

ͨͩ͠

Slide 61

Slide 61 text

• ͜ͷΑ͏ͳࢦඪ͸͋͘·Ͱ໨҆ • ࣮ࡍͷղऍ͸ݚڀऀࣗ਎ͷ੹೚Ͱ

Slide 62

Slide 62 text

༗ҙੑݕఆ

Slide 63

Slide 63 text

༗ҙੑݕఆ ॏͳΓͷେ͖͞͸Θ͔͚ͬͨͲɺ  ͜ͷࠩ͸ۮવʁ

Slide 64

Slide 64 text

༗ҙੑݕఆ ؍࡯͞Ε͕ͨࠩۮવੜͨ͡΋ͷͰ͋Δ Մೳੑʢ֬཰ʣΛܭࢉ͠·͠ΐ͏

Slide 65

Slide 65 text

t ݕఆ

Slide 66

Slide 66 text

t ݕఆ ← ฏۉͷࠩ ← ඪ४ภࠩ2ͷ࿨  1 2 2 2 1 2 1 − + − = n SD SD X X t ↑ ʢ֤܈ͷඪຊαΠζʣ ʢඪຊαΠζ͕౳͍͠৔߹ʣ ʢ٢ా, 1998, p. 186ʣ

Slide 67

Slide 67 text

͜Ε͖ͬ͞ݟͨʁ

Slide 68

Slide 68 text

pooled SD X X d 2 1 − = ←ɹฏۉͷࠩ ←ɹඪ४ภࠩ Cohen’s d ʮ͜Εʹ n Λ଍͢ͱ t ͬΆ͍ʂʯ

Slide 69

Slide 69 text

pooled SD X X d 2 1 − = 1 2 2 2 1 2 1 − + − = n SD SD X X t ʮt ͸ɺޮՌྔʹඪຊαΠζΛՃຯͨ͠΋ͷʯ

Slide 70

Slide 70 text

←ɹখ͍͞ํ͕ྑ͍ ←ɹେ͖͍ํ͕ྑ͍ t ஋͕େ͖͘ͳΔʹ͸ 1 2 2 2 1 2 1 − + − = n SD SD X X t ↑ɹେ͖͍ํ͕ྑ͍

Slide 71

Slide 71 text

n 1 2 3 4 5 ྟք஋ ྆ଆݕఆ5% 12.71 4.3 3.18 2.78 2.57 n 10 20 50 100 200 ྟք஋ ྆ଆݕఆ5% 2.23 2.09 2.01 1.98 1.97 *͜ΕΑΓେ͖͍਺஋ͩͬͨΒۮવͰͳ͍ͱΈͳ͢ t ͷྟք஋*

Slide 72

Slide 72 text

1 2 2 2 1 2 1 − + − = n SD SD X X t *2܈Ͱ n ͕ҟͳΔͱ͖ͷܭࢉ͸΋   ͏গ͠ෳࡶʹͳΓ·͢ * | 52.1-57.1| = √(15.32 + 16.72) / (30 - 1) ܭࢉͯ͠ΈΑ͏ Group A Group B ฏۉ஋ 52.1 57.1 ඪ४ภࠩ** 15.3 16.7 **ਪଌ౷ܭʢ༗ҙੑݕఆʣͰ࢖͏ͷ͸  ෆภ෼ࢄʹجͮ͘ඪ४ภࠩɻn ͷ  ୅ΘΓʹ n–1 Λܭࢉʹ࢖͍·͢ɻ

Slide 73

Slide 73 text

1 2 2 2 1 2 1 − + − = n SD SD X X t * 5 = 4.21 ܭࢉͯ͠ΈΑ͏ = 1.18 Group A Group B ฏۉ஋ 52.1 57.1 ඪ४ภࠩ** 15.3 16.7 *2܈Ͱ n ͕ҟͳΔͱ͖ͷܭࢉ͸΋   ͏গ͠ෳࡶʹͳΓ·͢ **ਪଌ౷ܭʢ༗ҙੑݕఆʣͰ࢖͏ͷ͸  ෆภ෼ࢄʹجͮ͘ඪ४ภࠩɻn ͷ  ୅ΘΓʹ n–1 Λܭࢉʹ࢖͍·͢ɻ

Slide 74

Slide 74 text

n 1 2 3 4 5 ྟք஋ ྆ଆݕఆ5% 12.71 4.3 3.18 2.78 2.57 n 10 20 50 100 200 ྟք஋ ྆ଆݕఆ5% 2.23 2.09 2.01 1.98 1.97 t ͷྟք஋ t = 1.18 ͸༗ҙͰͳ͍

Slide 75

Slide 75 text

͜͜·Ͱͷ·ͱΊ

Slide 76

Slide 76 text

• ޮՌྔ Cohen’s d • 2ͭͷάϧʔϓؒͷࠩΛඪ४Խͨ͠΋ͷ • t ݕఆ • ޮՌྔʹඪຊޡࠩͷӨڹΛՃຯͯ͠ɺͦͷ͕ࠩ ۮવ؍࡯͞ΕΔ֬཰Λࣔͨ͠΋ͷ • ݕఆ౷ܭྔ = ޮՌͷେ͖͞ x ඪຊͷେ͖͞ ʢೆ෩ݪ, 2002, p. 163ʣ

Slide 77

Slide 77 text

• Cohen’s d ͷ஥ؒ: • Hedges’ g • ෼฼ʹ฼ूஂͷඪ४ภࠩʢෆภ෼ࢄʹ جͮ͘ඪ४ภࠩʣΛ࢖͏ • Glass’ ⊿ • ෼฼ʹ౷੍܈ͷඪ४ภࠩΛ࢖͏

Slide 78

Slide 78 text

ؔ܎ͷڧ͞Λද͢ࢦඪ Pearson’s r / r2

Slide 79

Slide 79 text

• ม਺ؒͷؔ܎ͷେ͖͞Λද͢ • ࠷େ: 1.0ʢઈର஋ʣ • ࠷খ: 0 • ϐΞιϯͷੵ཰૬ؔ܎਺ r • r2 ʢ෼ࢄઆ໌཰ʣ Pearson’s r / r2

Slide 80

Slide 80 text

෼ࢄ෼ੳͷ৔߹

Slide 81

Slide 81 text

෼ࢄ෼ੳͷ৔߹ ௐ΂͍ͨཁҼͷ෼ࢄ η2 = ૯෼ࢄ SSA = SST

Slide 82

Slide 82 text

ҰཁҼ෼ࢄ෼ੳ SS df MS F p η2 A 847.609 2 423.805 0.955 .389 .022 Error 37259.655 84 443.567 Total 38107.264 MacR ʹ෇ଐͷσʔλΛ࢖ͬͯ MacR Ͱܭࢉ

Slide 83

Slide 83 text

ҰཁҼ෼ࢄ෼ੳ SS df MS F p η2 A 847.609 2 423.805 0.955 .389 .022 Error 37259.655 84 443.567 Total 38107.264 / = / = MS = SS / df

Slide 84

Slide 84 text

ҰཁҼ෼ࢄ෼ੳ SS df MS F p η2 A 847.609 2 423.805 0.955 .389 .022 Error 37259.655 84 443.567 Total 38107.264 / = ↑ɹඪຊαΠζ͕େ͖͍ͱ F ஋͕େ͖͘ͳΔ F = MSA / MSE = 423.805 / 443.567 = 0.955

Slide 85

Slide 85 text

ҰཁҼ෼ࢄ෼ੳ SS df MS F p η2 A 847.609 2 423.805 0.955 .389 .022 Error 37259.655 84 443.567 Total 38107.264 + = η2 = SSA / SST = 847.609 / 38107.264 = .022

Slide 86

Slide 86 text

ޮՌྔͷղऍ

Slide 87

Slide 87 text

• small: η2 = .01 • medium: η2 = .06 • large: η2 = .14 ਫຊɾ஛಺ (2008) • ͜ͷΑ͏ͳࢦඪ͸͋͘·Ͱ໨҆ • ࣮ࡍͷղऍ͸ݚڀऀࣗ਎ͷ੹೚Ͱ

Slide 88

Slide 88 text

͜͜·Ͱͷ·ͱΊ

Slide 89

Slide 89 text

• r ଒ͷޮՌྔ • ม਺ؒͷؔ܎ͷڧ͞Λ਺஋Ͱࣔͨ͠΋ͷ • ࠷େͰ 1.0ɺ࠷খͰ 0 • ෼ࢄ෼ੳͰ࢖͏ η2 ͸ r2 ͱࣅͨײ͡ • F ͱ η2 ͷҧ͍͸ඪຊαΠζΛߟྀ͢Δ͔Ͳ͏͔ • ݕఆ౷ܭྔ = ޮՌͷେ͖͞ x ඪຊͷେ͖͞ ʢೆ෩ݪ, 2002, p. 163ʣ

Slide 90

Slide 90 text

• η2 ͷ஥ؒ: • partial η2 • ෼฼ʹ SSA + SSE Λ࢖͏ • ω2 • ฼෼ࢄਪఆͷͨΊͷόΠΞεΛऔΓআ ͍ͨ΋ͷ

Slide 91

Slide 91 text

ޮՌྔͷ৴པ۠ؒ

Slide 92

Slide 92 text

ޮՌྔͷ৴པ۠ؒ • ฼ूஂͷਪఆΛલఏͱ͢ΔޮՌྔ͸఺ਪఆ • ඪຊޡࠩΛߟྀ͢Δඞཁ͕͋Δ • ޮՌྔͷ۠ؒਪఆ΋ߦ͏ͱΑ͍

Slide 93

Slide 93 text

• ఺ਪఆ • ฼ूஂΛ୅ද͢Δ఺ͷਪఆ • ۠ؒਪఆ • ఺ਪఆͷ͹Β͖ͭͷਪఆ • ৴པ۠ؒ • ͋Δ֬཰ʢe.g., 95%ʣͰਪఆ͞Εͨ۠ؒ ޮՌྔͷ৴པ۠ؒ

Slide 94

Slide 94 text

࣮ࡍʹ ܭࢉͯ͠Έ·͠ΐ͏

Slide 95

Slide 95 text

http://www.mizumot.com/stats/eﬀectsize.xls ޮՌྔͷܭࢉγʔτ

Slide 96

Slide 96 text

http://www.mizumot.com/stats/eﬀectsize.xls

Slide 97

Slide 97 text

Cohen’s d Λܭࢉͯ͠ΈΑ͏

Slide 98

Slide 98 text

Cohen’s d Λܭࢉͯ͠ΈΑ͏

Slide 99

Slide 99 text

Cohen’s d Λܭࢉͯ͠ΈΑ͏

Slide 100

Slide 100 text

Cohen’s d Λܭࢉͯ͠ΈΑ͏

Slide 101

Slide 101 text

η2 Λܭࢉͯ͠ΈΑ͏

Slide 102

Slide 102 text

η2 Λܭࢉͯ͠ΈΑ͏

Slide 103

Slide 103 text

http://www.langtest.jp Eﬀect Size Calculator @

Slide 104

Slide 104 text

• σʔλ෼ੳ͕ͳΜͰ΋  Ͱ͖ͪΌ͏΋ͷ͍͢͝  ΢ΣϒΞϓϦ • ޮՌྔ d, g ͱͦͷ৴པ۠ ؒΛܭࢉͯ͘͠ΕΔ • ਫຊಞ͞Μʢؔ੢େֶʣ ͕։ൃ͠ɺແྉͰެ։ • ͓ྱ͸Ϗʔϧ·ͨ͸  νϣίϨʔτͰ

Slide 105

Slide 105 text

No content

Slide 106

Slide 106 text

No content

Slide 107

Slide 107 text

No content

Slide 108

Slide 108 text

No content

Slide 109

Slide 109 text

ࢀߟจݙ

Slide 110

Slide 110 text

ࢀߟจݙ • ӳޠڭࢣͷͨΊͷڭҭσʔλ ෼ੳೖ໳ • ༗ҙੑݕఆͷ͘͠Έ΍ͦͷݶ քʹ͍ͭͯ΋ղઆ

Slide 111

Slide 111 text

ࢀߟจݙ • ຊ౰ʹΘ͔Γ΍͍͘͢͢͝େ ੾ͳ͜ͱ͕ॻ͍ͯΔ͘͝ॳา ͷ౷ܭͷຊ • େ੾ͳ͜ͱΛ਺ࣜΛަ͑ͯஸ ೡʹղઆ

Slide 112

Slide 112 text

ࢀߟจݙ • ֎ࠃޠڭҭݚڀϋϯυϒοΫ • هड़౷ܭɺਪଌ౷ܭɺޮՌྔ ΋ؚΊͯ໢ཏతͳҰ࡭

Slide 113

Slide 113 text

ࢀߟจݙ • ఻͑ΔͨΊͷ৺ཧ౷ܭ • ޮՌྔʹ͍ͭͯษڧ͢ΔͳΒ ඞಡ

Slide 114

Slide 114 text

ࢀߟจݙ • ͸͡Ίͯͷӳޠڭҭݚڀ • ݚڀͷೖޱΛղઆ͢ΔҰ࡭ɻ ࠓ೔ͷ಺༰͸ୈ6ষΛิ଍͢ Δ΋ͷ

Slide 115

Slide 115 text

1. ෼෍ͱඪ४ภࠩ 2. ༗ҙੑʢݕఆʣͱޮՌྔ 3. ޮՌྔͷछྨ • ࠩͷେ͖͞Λද͢ d ଒ • ؔ܎ͷڧ͞Λද͢ r ଒ 4. ޮՌྔͷ৴པ۠ؒ 5. ޮՌྔͷܭࢉ Ken Urano [email protected] http://bit.ly/let2016ws ྔతσʔλΛѻ͏ͱ͖ʹ ஌͓͖͍ͬͯͨ͜ͱ —σʔλͷ෼෍ɺ༗ҙੑɺޮՌྔɺ৴པ۠ؒ—

Slide 116

Slide 116 text

ࢀߟจݙ • Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Lawrence Earlbaum Associates. • ೆ෩ݪே࿨. (2002). ʰ৺ཧ౷ܭֶͷجૅ: ౷߹తཧղͷͨΊʹʱ౦ژ: ༗൹ֳ. • લాܒ࿕ɾࢁ৿ޫཅ (ฤ). (2004). ʰӳޠڭࢣͷͨΊͷڭҭσʔλ෼ੳೖ໳: तۀ͕มΘ ΔςετɾධՁɾݚڀʱ౦ژ: େमؗॻళ. • ਫຊಞɾ஛಺ཧ. (2008). ʮݚڀ࿦จʹ͓͚ΔޮՌྔͷใࠂͷͨΊʹ: جૅత֓೦ͱ஫ҙ ఺ʯʰӳޠڭҭݚڀʱୈ31߸, 57–66. Retrieved from http://www.mizumot.com/files/ EﬀectSize_KELES31.pdf • େٱอ֗ѥɾԬాݠհ. (2012). ʰ఻͑ΔͨΊͷ৺ཧ౷ܭ: ޮՌྔɾ৴པ۠ؒɾݕఆྗʱ ౦ژ: Ⴛ૲ॻ๪. • Plonsly, L., & Oswald, F. (2014). How big is “big”? Interpreting eﬀect Sizes in L2 research. Language Learning, 64, 878–912. doi:10.1111/lang.12079 • ஛಺ཧɾਫຊಞ (ฤ). (2014). ʰ֎ࠃޠڭҭݚڀϋϯυϒοΫ: ݚڀख๏ͷΑΓྑ͍ཧղ ͷͨΊʹ (վగ൛)ʱ౦ژ: দദࣾ. • Ӝ໺ݚɾ࿱ཧཅҰɾాத෢෉ɾ౻ా୎࿠ɾ∁໦ѥرࢠɾञҪӳथ. (2016). ʰ͸͡Ίͯͷ ӳޠڭҭݚڀ: ԡ͓͖͍͑ͯͨ͞ίπͱϙΠϯτʱ౦ژ: ݚڀࣾ. • ٢ాण෉. (1998). ʰຊ౰ʹΘ͔Γ΍͍͘͢͢͝େ੾ͳ͜ͱ͕ॻ͍ͯ͋Δ͘͝ॳาͷ౷ܭ ͷຊʱژ౎: ๺େ࿏ॻ๪.