医学统计

医学统计、生物统计基本概念与方法.

304
0

美国高校统计学(含生物统计学)专业的排名比较知名的是US News Best Graduate Schools排名,刚看了下好像仍是2010的排名,3年过去了,不知新排名会如何。

201310231427547063.jpg

不过这种排名一般来讲变化不会太大,前几位的总是那几所学校,斯坦福、加州伯克利、哈佛等等。

另外一个颇具影响力的排名是Top Universities的世界排名(QS世界大学排名),我第1次看到,亚洲如新加坡国立大学、香港科技大学在统计学排名中如此靠前,我们的清华大学排名第11位,上海交大排第18,我印象中人大的统计也非常牛,不知为何没在前20位。

201310231440462501.jpg

3.3k
0

|Rank | School Name | Address | Score|

|---- | -------------- | ------ | ----|

|#1 in Statistics | Stanford University (Department of Statistics) | Stanford, CA | 5.0|

|#2 in Statistics | University of California--Berkeley (Department of Statistics) | Berkeley, CA | 4.7|

|#3 in Statistics (tie) | Harvard University (Department of Biostatistics) | Boston, MA | 4.6|

|#3 in Statistics (tie) | Johns Hopkins University (Department of Biostatistics) | Baltimore, MD | 4.6|

|#3 in Statistics (tie) | University of Washington (Department of Biostatistics) | Seattle, WA | 4.6|

|#6 in Statistics (tie) | Harvard University (Department of Statistics) | Cambridge, MA | 4.4|

|#6 in Statistics (tie) | University of Chicago (Department of Statistics) | Chicago, IL | 4.4|

|#8 in Statistics (tie) | Carnegie Mellon University (Department of Statistics) | Pittsburgh, PA | 4.3|

|#8 in Statistics (tie) | University of North Carolina--Chapel Hill (Department of Biostatistics) | Chapel Hill, NC | 4.3|

|#8 in Statistics (tie) | University of Washington (Department of Statistics) | Seattle, WA | 4.3|

|#11 in Statistics | University of Michigan--Ann Arbor (Department of Biostatistics) | Ann Arbor, MI | 4.2|

|#12 in Statistics (tie) | Duke University (Department of Statistical Science) | Durham, NC | 4.1|

|#12 in Statistics (tie) | University of California--Berkeley (Group in Biostatistics) | Berkeley, CA | 4.1|

|#12 in Statistics (tie) | University of Michigan--Ann Arbor (Department of Statistics) | Ann Arbor, MI | 4.1|

|#12 in Statistics (tie) | University of Pennsylvania (Department of Statistics) | Philadelphia, PA | 4.1|

|#16 in Statistics (tie) | Columbia University (Department of Statistics) | New York, NY | 4.0|

|#16 in Statistics (tie) | North Carolina State University (Department of Statistics) | Raleigh, NC | 4.0|

|#16 in Statistics (tie) | University of Wisconsin--Madison (Department of Statistics) | Madison, WI | 4.0|

|#19 in Statistics | University of North Carolina--Chapel Hill (Department of Statistics & Operations Research) | Chapel Hill, NC | 3.9|

|#20 in Statistics (tie) | Cornell University (Department of Statistical Science) | Ithaca, NY | 3.8|

|#20 in Statistics (tie) | Iowa State University (Department of Statistics) | Ames, IA | 3.8|

|#20 in Statistics (tie) | Pennsylvania State University (Department of Statistics) | University Park, PA | 3.8|

|#20 in Statistics (tie) | Texas A&M University--College Station (Department of Statistics) | College Station, TX | 3.8|

|#24 in Statistics (tie) | University of Minnesota--Twin Cities (School of Public Health) | Minneapolis, MN | 3.7|

|#24 in Statistics (tie) | University of Minnesota--Twin Cities (School of Statistics) | Minneapolis, MN | 3.7|

|#24 in Statistics (tie) | University of Wisconsin--Madison (School of Medicine and Public Health) | Madison, WI | 3.7|

|#27 in Statistics (tie) | Columbia University (Department of Biostatistics) | New York, NY | 3.6|

|#27 in Statistics (tie) | Purdue University--West Lafayette (Department of Statistics) | West Lafayette, IN | 3.6|

|#27 in Statistics (tie) | University of California--Los Angeles (Department of Biostatistics) | Los Angeles, CA | 3.6|

|#27 in Statistics (tie) | University of Texas MD Anderson (Department of Biostatistics) | Houston, TX | 3.6|

|#31 in Statistics (tie) | Johns Hopkins University (Department of Applied Mathematics and Statistics) | Baltimore, MD | 3.5|

|#31 in Statistics (tie) | University of California--Davis (Department of Statistics) | Davis, CA | 3.5|

|#31 in Statistics (tie) | University of California--Los Angeles (Department of Statistics) | Los Angeles, CA | 3.5|

|#31 in Statistics (tie) | University of Pennsylvania (Department of Biostatistics & Epidemiology) | Philadelphia, PA | 3.5|

|#31 in Statistics (tie) | Yale University (Department of Biostatistics) | New Haven, CT | 3.5|

|#31 in Statistics (tie) | Yale University (Department of Statistics) | New Haven, CT | 3.5|

|#37 in Statistics (tie) | Emory University (Department of Biostatistics and Bioinformatics) | Atlanta, GA | 3.4|

|#37 in Statistics (tie) | Ohio State University (Department of Statistics) | Columbus, OH | 3.4|

|#37 in Statistics (tie) | University of Illinois--Urbana-Champaign (Department of Statistics) | Champaign, IL | 3.4|

|#40 in Statistics (tie) | Rutgers University--New Brunswick (Department of Statistics and Biostatistics) | Piscataway, NJ | 3.3|

|#40 in Statistics (tie) | University of Florida (Department of Statistics) | Gainesville, FL | 3.3|

|#40 in Statistics (tie) | University of Iowa (Department of Statistics and Actuarial Science) | Iowa City, IA | 3.3|

|#43 in Statistics | Rice University (Department of Statistics) | Houston, TX | 3.2|

|#44 in Statistics (tie) | Brown University (Department of Biostatistics) | Providence, RI | 3.1|

|#44 in Statistics (tie) | Colorado State University (Department of Statistics) | Fort Collins, CO | 3.1|

|#44 in Statistics (tie) | Duke University (Department of Biostatistics and Bioinformatics) | Durham, NC | 3.1|

|#44 in Statistics (tie) | Florida State University (Department of Statistics) | Tallahassee, FL | 3.1|

|#44 in Statistics (tie) | University of Connecticut (Department of Statistics) | Storrs, CT | 3.1|

|#44 in Statistics (tie) | Vanderbilt University (Department of Biostatistics) | Nashville, TN | 3.1|

|#50 in Statistics (tie) | Boston University (School of Public Health) | Boston, MA | 3.0|

注:前50名。

来源:https://www.usnews.com/best-graduate-schools/top-science-schools/statistics-rankings

[1]: https://www.cnstat.org/community/usr/uploads/2020/04/3422259443.png

1.8k
0
StatX 发表于 2019-03-19 7:45 am

2019年6月20日发布:JCR影响因子-2018官方完整版

https://www.cnstat.org/community/sites/old/uploads/2019/06/2808775731.xlsx

2018年发布:2018期刊影响因子.zip

https://www.cnstat.org/community/sites/old/uploads/2019/03/678602978.zip

2017年发布:2017期刊影响因子Journal+Citation+Reports.zip

https://www.cnstat.org/community/sites/old/uploads/2019/03/963197951.zip

大类分区表:2014年SCI杂志JCR分区大类一区表.zip

https://www.cnstat.org/community/sites/old/uploads/2019/03/4232402929.zip


516
0

[细说统计]之-统计基础(2)

统计学中的假设检验,是一种基于概率的反证法,我们称之为“小概率反证法”。与数学上的反证法不同,用假设检验的方法证明了的命题,也有可能是错误的。

用假设检验的方法,证明小明同学作弊了,需要分三步走

Step 1:明确两个假设(命题),并明确冤枉小明的概率大小

H0(无效假设):小明没有作弊
H1(备择假设):小明作弊了
检验水准:α=0.05(当拒绝H0、接受H1时,犯错的概率,即冤枉小明的概率)

Step 2:收集小明没有作弊的信息,计算没有作弊的概率

既然是反证法,当然要基于小明没有作弊(即H0)去做推演
在这一步骤中,需要根据收集的信息(样本数据),计算出统计量的大小,再根据统计量的分布,求出相应的P值-小明没有作弊的概率。

比如,我们计算出统计量t的值,利用t分布就可求得相应的P值。

当然,有时概率P可以直接计算,如Fisher确切概率法,就没有统计量。​

Step 3:根据统计量对应的P值作出推断

在数学上,从H0出发进行推演,若证明H0这个命题是假的,则证明了H1为真,因为H0与H1互相对立,其中必有一真一假。而假设检验,则是根据P≤α 则拒绝H0、接受H1这个规则,进行H0与H1真伪的判断。

比如,根据统计量t的值,我们得到的P值为0.05,这时我们就可以下结论:小明作弊了(拒绝H0、接受H1)。这里的逻辑就是:这个概率P是基于H0-小明没有作弊得到的,既然小明没有作弊的概率这么小,那么我们这时就得相信:嗯,H0够假、H1够真,所以小明作弊了。但理论上,H0并不是100%错。

这就是利用假设检验的方法,证明小明作弊与否的步骤。

要是得到的P值大于α--“小明没有作弊”这个命题还不够那么假,那我们就不能拒绝H0。不拒绝H0,我们也不会接受H0,不会认为他没作弊。Why?因为我们的目的,就是要证明他作弊了(假设检验就是要证明H1为真)。现在的情况是没能证明他作弊,可能是我们掌握的信息量还是太小。只要掌握的信息量足够大,我们就一定能用假设检验的方法证明他作弊了:-),因为绝大多数情况下,P值会随着信息量的增大而减小。


在假设检验过程中,α就像一把尺子,让我们做出推断:

-- 若统计量对应的p值≤α,我们的推断就是:H0为假、H1为真。

虽然α很小(常用0.05,这就是统计中小概率事件的概率水平--我们认为在一次抽样中,小概率事件不太可能发生),但毕竟这个α不是0,多次基于α这把尺子,拒绝H0(接受H1为真),犯错率就是α。当α=0.05时,我们拒绝H0若有100次,其中就会有5次是错的,即H0为真但被拒绝了-小明被冤枉了。

这就是统计学中做证明题的方式:​假设检验方法,用一个小概率值α作为判断命题真伪的标尺,这把尺子多数情况下Ok,少数情况会犯错。所谓,常在河边走, 哪有不湿鞋?这把尺子本身所具有的特性,就决定了用它进行度量的结果。因为它是冤枉小明的概率大小,对于我们而言,不冤枉一个好人,远比不放过一个坏人更重要,所以,假设检验要规定α而且它的值要很小。至于为什么α常用0.05,或0.01这样的水平,则是一种约定俗成(当初老大拍脑门定了个0.05的概率水平,一直沿用至今。。。)


1.3k
0
StatX 发表于 2013-04-13 12:12 pm

有N多个新的开始,就有N多个不同的理由。

未来仍有很多不确定性,但方向却是唯一的,那就是向前...

4.9k
0
StatX 发表于 2017-02-28 1:54 pm

Finite versus Infinite Populations

从概念上,无限总体指总体中的个体数是无限的,当然没有绝对的无限,是相对而言,数量足够大而已;有限总体则是指总体中的个体数是有限的或固定的。

区别总体的“无限”或“有限”,除了数量上的定义,也可以从样本角度去考虑:如果抽样研究(非放回)的样本量对于总体的个体数影响很小或几乎无影响,就可以认为总体是“无限”的。

在实际应用中,需要区别无限总体与有限总体吗?

Y. 因为某些标准误的计算是基于:总体无限大或放回抽样。

有限总体校正(Finite Population Correction)

当样本量(n)较大(相对于总体N而言),进行非放回抽样研究时,计算均数标准误或率的标准误时:

SE = [(standard deviation)/sqrt(n)]*fpc , 其中 fpc=sqrt[(N ‐ n)/(N ‐ 1)]

5.7k
0

前几天一医生朋友给我发信息,他的文章(SCI)编辑问为什么有的地方用卡方,有的地方却用的Fisher的方法。

(估计也不是什么太大牌的杂志,否则不会问出这个问题)

统计内容我帮他做的,自然也要帮人帮到底,遂从wikipedia[1]上抄了几句,后面的稍做了修改:

Fisher's exact test is a statistical significance test used in the analysis of contingency tables. Although in practice it is employed when sample sizes are small, it is valid for all sample sizes.

With large samples, a chi-squared test can be a good choice, but the significance value it provides is only an approximation.

The approximation is inadequate when sample sizes are small, or the data are very unequally distributed among the cells of the table, resulting in the cell counts predicted on the null hypothesis (the “expected values”) being low. The usual rule of thumb for deciding whether the chi-squared approximation is good enough is that the chi-squared test is not suitable when the expected values in 20% or more of the cells of a contingency table are below 5.

可能有人会问,为什么不用卡方的连续性校正?统计课本里就这么教的呀。

事实上,卡方的连续性校正方法(Yates's correction for continuity, 1934),多数情况下矫枉过正。而Fisher's exact test是更好的选择,只不过以前限于计算能力,这个“exact test”的方法不容易实现,而现在则不存在这个障碍。

看一个使用卡方连续性校正结论发生反转的例子:

所以当卡方检验的P值处于α(一般为0.05)附近时,需要注意卡方检验的近似程度是否满足要求,必要时使用Fisher精确概率法替代卡方检验。

[1] https://en.wikipedia.org/wiki/Fisher's_exact_test


   13
   21
   0
浏览设置
隐藏帖子详情
启用无限滚动载入
前一页
12
下一页
该主题下的所有帖子将 已删除 ?
待审的草稿 ... 点击恢复编辑
放弃草稿