hellogѸ˥֥               ǿ     helhub (The HEL Hub) 2011-03     ڡ     ɽ    

hellogѸ˥֥ / 2011-03-24

01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

2011-03-24 Thu

#696. Log-Likelihood Test [corpus][bnc][statistics][lltest]

[2010-03-04-1]εǿ줿ѥؤǤϳƼ׼ˡѤ롥ĤˡΤʤǤ⡤ɽΥѥ֤٤Ӥꡤcollocation ٹ礤¬Τ˹ѤƤΤ Log-Likelihood Test ( LL Test, G Test, G2 Test ʤɤȤ˸ƤФ븡Ǥ롥ѥθ줿ʤΤǥΰۤʤ륳ѥ֤ǤӤǽǤꡤƱŪǰˤ褯ѤƤ2踡 ( Chi-Squared Test ) ⤤ĤǤ줿ˡɾƤꡤǶΥѥǤϹѤƤ롥㤨С2踡ϴ٤5꾯ʤȤٸ򰷤Ȥѥ礭ΤȾΤӤȤ˿㤯ʤ뤬Log-Likelihood Test Ϥαƶˤ [ Rayson and Garside 2 ]
Log-Likelihood Test δŪʹͤϡѥȤˤɽδԤи١ʴ١ˤФͤȼºݤ˽и١ʴѻ١ˤκñʸȹͤۤɤ˶Ƥ뤫ɤȽꤹȤΤǤ롥ȤơΤ褦ʥǥBNC ( The British National Corpus ) äե֥ѥȽ񤭸ե֥ѥ̤ξ֥ѥ֤ f*ck Ȥ four-letter word ٤Ӥ롥BNCweb ꤳΥɤ򸡺ȡΤ褦ʷ̤줿

CategoryNo. of wordsNo. of hitsDispersion (over files)Frequency per million words
Spoken10,409,85857963/90855.62
Written87,903,571743172/3,1408.45
total98,313,4291,322235/4,04813.45


׽ۤɤޤǤʤDZ "Frequency per million words" 򸫤Сf*ck Ūäդ¿Ѥ뤳Ȥʬ뤬ϤŪ΢դ롥ޤ̵Ȥơäե֥ѥȽ񤭸ե֥ѥδ֤Ǥ f*ck ٺϸϰǤꡤθ˴ؤξԤ˰̣Τ뺹Ϥʤפꤹ롥Ωϡäե֥ѥȽ񤭸ե֥ѥδ֤Ǥ f*ck ٺϸϰǤʤθ˴ؤξԤκϰ̣פȤʤ롥̵⤬ٻ뤫ɤΤŪǤ롥

 Corpus 1Corpus 2Total
Frequency of wordaba+b
Frequency of other wordsc-ad-bc+d-a-b
Totalcdc+d


Log-Likelihood Test Ѥ Log-Likelihood ratio пפϡɽΤdzƥ֥ѥ ( c, d ) ȡƥ֥ѥǤ f*ck ٿ ( a, b ) ʬɽˤޤȤ᤿ǡ줾δ E1 E2 򲼤 (1) μǵᡤͤ (2) μƵ롥

(1) E1 = c*(a+b)/(c+d); E2 = d*(a+b)/(c+d)
(2) LL = 2*((a*log(a/E1))+(b*log(b/E2)))

f*ck οͤǷ׻ȡʲΤ褦ˤʤ롥

E1 = 10409858*(579+743)/(10409858+87903571) = 139.979170861796
E2 = 87903571*(579+743)/(10409858+87903571) = 1182.0208291382
LL = 2*((579*log(579/139.979170861796))+(743*log(743/1182.0208291382))) = 954.2115

Log-likelihood ratio Ȥ 954.2115 ȤͤФ롥ˤͤŬڤͭտ̾ 5%, 1%, 0.1%ˤб륫ͤӤ롥2 * 2 ʬɽФ׻Ǥϼͳ1ΥͤѤ뤳ȤˤʤäƤꡤͤͭտ 5%, 1%, 0.1% νˤ줾 3.84, 6.63, 10.83 Ǥ롥954.2115 Log-Likelihood ratio ͭտ 0.1% б 10.83 ⤺äȹ⤤Τǡ0.1% ͭտǵ̵ϴѤ롥СŪˤϵ̵⤬ǤΨ 0.1% ˤޤȹͤƤ褤ȤȤǤ롥Τ褦ˤΩäե֥ѥȽ񤭸ե֥ѥδ֤Ǥ f*ck ٺϸϰǤʤθ˴ؤξԤκϰ̣פ򤵤뤳Ȥˤʤ롥
Log-Likelihood Test ϰʾΤ褦˿ʤ뤬θԤʤˤäƤΤäƤɬפ롥̤ˤϡ׻٤ 5 򲼲륻뤬1ĤǤ⤢ˤϡ٤Ȥ롥 the Cochran rule ȸƤФƤ뤬꤭٤ʥ롼󵯤 Rayson, Berridge, and Francis (8) ˤС٤٤ͤͭտ 5% 13 1% 11 0.1% 8 Ȥͭտ 0.01% ꤹд 1 ˤѤ٤ΤǡRayson et al. ϥѥؤǴŪѤƤ3Ĥο˲äơ0.01% οб륫ͤ 15.13 ˤޤǤθ侩Ƥ롥
פˤϾܤʤɽ 2ʥ֡˥ѥ֤ǤӤȤǴñѤ뤳ȤǤ븡ȤơLog-Likelihood Test αϰϤϹ׻Τ Rayson Log-likelihood calculator ʤɤǤФ褤ܵϤΥڡεҤȥʸ򻲹ͤˤˡ
BNC Ѥ f*ck ϢʬۤθϡMcEnery et al. (264--86) Υǥ˾ܤ
ϢơϹԤʤʤäĤܥ֥ǰä gorgeous Ĵ ([2010-08-16-1], [2010-08-17-1],[2010-12-25-1]) ʤɤ⻲ȡ

Rayson, P., D. Berridge , and B. Francis. "Extending the Cochran Rule for the Comparison of Word Frequencies between Corpora." Le poids des mots: Proceedings of the 7th International Conference on Statistical Analysis of Textual Data (JADT 2004), Louvain-la-Neuve, Belgium, March 10-12, 2004. Ed. Purnelle G., Fairon C., and Dister A. Louvain: Presses universitaires de Louvain, 2004. 926--36. Available online at http://www.comp.lancs.ac.uk/computing/users/paul/publications/rbf04_jadt.pdf .
Rayson, P. and R. Garside. "Comparing Corpora Using Frequency Profiling". Proceedings of the Workshop on Comparing Corpora, Held in Conjunction with the 38th Annual Meeting of the Association for Computational Linguistics (ACL 2000), 1-8 October 2000, Hong Kong. 2000. 1--6. Available online at http://www.comp.lancs.ac.uk/computing/users/paul/phd/phd2003.pdf .
McEnery, Tony, Richard Xiao, and Yukio Tono. Corpus-Based Language Studies: An Advanced Resource Book. London: Routledge, 2006.

[ | ѥڡ ]

2026 : 01 02 03 04 05 06 07 08 09 10 11 12
2025 : 01 02 03 04 05 06 07 08 09 10 11 12
2024 : 01 02 03 04 05 06 07 08 09 10 11 12
2023 : 01 02 03 04 05 06 07 08 09 10 11 12
2022 : 01 02 03 04 05 06 07 08 09 10 11 12
2021 : 01 02 03 04 05 06 07 08 09 10 11 12
2020 : 01 02 03 04 05 06 07 08 09 10 11 12
2019 : 01 02 03 04 05 06 07 08 09 10 11 12
2018 : 01 02 03 04 05 06 07 08 09 10 11 12
2017 : 01 02 03 04 05 06 07 08 09 10 11 12
2016 : 01 02 03 04 05 06 07 08 09 10 11 12
2015 : 01 02 03 04 05 06 07 08 09 10 11 12
2014 : 01 02 03 04 05 06 07 08 09 10 11 12
2013 : 01 02 03 04 05 06 07 08 09 10 11 12
2012 : 01 02 03 04 05 06 07 08 09 10 11 12
2011 : 01 02 03 04 05 06 07 08 09 10 11 12
2010 : 01 02 03 04 05 06 07 08 09 10 11 12
2009 : 01 02 03 04 05 06 07 08 09 10 11 12

ǽ: 2026-06-26 03:46

Powered by WinChalow1.0rc4 based on chalow