（SAS Institute Japan株式会社／JMPジャパン事業部） 小野裕亮
このブログ記事のすべての責任は小野個人だけにあります．所属組織には一切の責任はありません．また，Theodore Porter先生およびStephen Stigler先生の許可を得て，両先生のメールを公開していますが，このメールを公開したことに伴うすベての責任も，小野個人一人だけにあります．
以前，次のようなブログ記事を書きました．
このブログ記事では，以下の2つの噂に対しての，Theodore Porter先生とStephen Stigler先生の意見とアドバイスを紹介します．両先生から，メール内容を公開してよいとの許可は得ています．
- 「統計学は科学の文法である」とKarl Pearsonはどこかで書き残しているのか？
- 「K. Pearson=記述統計学 vs R.A. Fisher=推測統計学」という対立図式は，米国でも語られているのか？
1についてはPorter先生だけに，2についてはPorter先生とStigler先生に尋ねました．とても貴重な意見とアドバイスだと思いますので，誤字なども修正せず原文そのままを以下に引用します．
このブログ記事を読むような方には説明不要だと思いますが，Theodore Porter先生とStephen Stigler先生は統計学史の専門家です．
Porter先生への私の質問メール
Dear Prof. Porter,
I am Yusuke Ono (Mr.), tester at SAS Institute Japan. I asked about your book review for the Lady Tasting Tea in 2016, and about probability in von Kries in 2019.
I am checking some kind of urban legends about Karl Pearson, but I completely stuck.
Could you please give me any advice for my following two questions if you are also interested in these urban legends?
(1) In American Statistical Association’s Twitter on Aug. 5th 2015, they “quoted" the word by Karl Pearson as below.
“Statistics is the grammar of science.” — Karl Pearson
#statquotes
The link of this tweet is https://twitter.com/AmstatNews/status/628658305167134720
As far as I search, Prof. Sastry G. Pantula also said “As Karl Pearson wrote, “statistics is the grammar of sicence."” in AmStat News March 2008.
The link of the newsletter is https://magazine.amstat.org/wp-content/uploads/2008an/march08.pdf
I have searched the origin of this word for more than two years, but I still cannot find it.
I know “Grammar of Science” is the book title by Karl Pearson, and as you wrote “THE STATISTICAL GRAMMAR” (on pp.211) section in your book, Karl Pearson: The Scientific Life in a Statistical Age, there must be a strong connection between K.Pearson’s scientific methods and statistics. But I cannot find a word like “Statistics is the grammar of science” or “Statistics is the one of the grammars of science” in Karl Pearson’s writings. If you know the origin of this quote, I would like you to tell it to me.
(2) In some textbooks for statistics, statistics is classified into “descriptive statistics” and “inferential statistics”. In some Japanese literatures, statistics by Karl Pearson is said as the old “descriptive statistics”, and statistics by R.A. Fisher is said as the new “inferential statistics”.
Is this labeling for Karl Pearson and R.A. Fisher also popular in the U.S.?
I know that in Grammar of Science, Karl Pearson claims that “That all science is description and not explanation”. But this is a contrast between “description” and “explanation”, not a contrast between “description” and “inference”, and he developed and used many inferential statistical methods in his papers.
Best Regards,
Yusuke Ono (Mr.)
JMP Japan Group
SAS Institute Japan
Porter先生からの返信メール
Dear Yusele Onoe, I will say first that Stephen is a betterer authority than I am on your second question, but I think you are quite correct in what you say--indeed you have hit the nail on the head (as we say). Fisher I think developed a sense of statistical inference that went well beyond Pearson, but at the same time he [ushed new forms of statistical inference that were largely unfamiliar to the statistics organizations like the ISO.
As for the identity of statistics with scientific method, I do not believe that he ever said this, though he certainly did come to see statistics as fundamental to science as a method. We must not forget however that he wrote the Grammar of Science just BEFORE his turn to statistics, and he definitely was not thinking of statistics in his formative writings on scientific method. I would not exclude the possibility that he could have said "statistics is scientific method" or the like, but I have never seen this, and if he had said this in print, we could surely find it from a search.
I hope this is helpful,
Theodore Porter
Stigler先生への私の質問メール
Dear, Prof. Stigler
I am very sorry for this sudden e-mail. I am Yusuke Ono (Mr.), tester at SAS Institute Japan.
Although I am just an amateur, I am interested in “modern” statistical history as a person who works for statistical software.
I would like to ask a question. I have asked the same question to Prof. Porter, and he advised that you are better authority.
In some textbooks for statistics, statistics is classified into “descriptive statistics” and “inferential statistics”. In some Japanese literatures, statistics by Karl Pearson is said as the old “descriptive statistics”, and statistics by R.A. Fisher is said as the new “inferential statistics”.
Is this labeling for Karl Pearson and R.A. Fisher also popular in the U.S.?
I know that in Grammar of Science, Karl Pearson claims that “That all science is description and not explanation”. But this is a contrast between “description” and “explanation”, not a contrast between “description” and “inference”, and he developed and used many inferential statistical methods in his papers.
I think I understand that Fisher developed new interpretations for some existing methods. For example, he interpreted newly the posteriori distribution with uniform prior as “likelihood” in 1922 (just one hundred years ago). He also proved some statistics are distributed as t-distribution. He also suggested “fiducial” limits. These interpretations and methods must be Fisher’s works, but I feel it’s strange to use the contrast, “descriptive” v.s. “inferential” for K.Pearson v.s. Fisher.
I am also sorry for my poor English.
Best Regards,
Yusuke Ono (Mr.)
JMP Japan Group
SAS Institute Japan
Stigler先生からの返信メール
Dear Yusuke Ono,
That is not a labeling I am familiar with and I do not think it is particularly accurate. The relationship is more complex. Fisher built upon a framework Pearson had employed and both did inferential statistics. I attach two papers that probably are more than you are looking for but easy to send.
Regards,
Stephen Stigler
なお，返信メールで述べられている2つの論文は，以下の論文です．
Stigler, S. (2005) Fisher in 1921. Statistical Science. 20(1), 32-49.
Stigler, S. (2008) Karl Pearson’s Theoretical Errors and the Advances They Inspired. Statistical Science. 23(2), 261-271.
この後，Fisherおよびその周りの人々による最尤推定の議論については，以下の論文をすすめてくださいました．
Stigler, S. (2007) The Epic Story of Maximum Likelihood. Statistical Science. 22(4), 598-620.
