以往學者研究「中山記歷、養生記逍」兩記的真偽，都是從文學或歷史的角度作判斷。本文以統計觀點，從作者的寫作風格來分析。第二節文獻探討，就使用「浮生六記」的版本及數據作說明；第三節使用變異數分析、Kruskal-Wallis 無母數方法、Kolmogrov-Smirnov無母數方法、 迴歸分析診斷工具各種殘差等方法，對每句字數之特徵值及次數分配作比較；第四節以Bootstrap 方法建構虛字「之」出現次數之分配作比較；第五節利用等候時間( wait time ) 分配觀念對虛字之、矣、者、也在每一篇文章中每隔多少字出現作比較；第六節將每一篇文章作分割，以「之」出現次數為應變數，「矣、者、也」出現次數為自變數，使用計數迴歸分析，分別採用 Poisson 模式及負二項模式作比較。第七節分析與結論。

關鍵詞：Kruskal-Walis 秩檢定，K-S 相等分配檢定，迴歸診斷，自力更生方法，等候時間分配，Poisson 迴歸，負二項分配迴歸

In Shen SanPo’s “ Six Chapters of a Floating life “, “ Experience” and “ The way of life” was proved by many scholars to be counterfeit literature.  It was apparently insubstantial reasons such as some scholars were insistent that it was Shen SanPo’s works.

In this paper we find out the statistical evidence of the difference between “ Experience” and another of four chapters –wedded bliss, the little pleasures of life, sorrow, the joys of travel.  We use four kinds of literate character and a lot of statistical methods, such as

1. The distribution of wordage of each sentence to summarize data, we use ANOVA, Kruskal-Walis rank test, K-S equal distribution test, regression diagnosis .

2. We use bootstrap method to generate the sampling distribution of “之”, “也”, “矣”, “者”.

3. Compare wait-time distribution of the wordage between each  “之”, “也”, “矣”, “者” in every chapters.

4. We subdivide each chapter by 250 words, and produced relative location by 250i/totals wordage.  Let the location of “之” is dependent variable and the other word’s location are independence variables to formulate Poisson regression models and negative binomial distribution  regression model .

The results of analysis showed “Experience” other than from the other four chapters which was written by Shen SanPo.

Keywords：Kruskal-Walis rank test, K-S equal distribution test, regression diagnosis . bootstrap method, wait-time distribution, Poisson regression , negative binomial distribution  regression .