查看“︁零假设”︁的源代码

{{Expand language|1=en|time=2024-05-29T07:41:21+00:00}}
{{NoteTA
|G1=Math
|T=zh-tw:虛無假說;zh-hk:虛無假設;zh-cn:零假设
|2=zh-tw:對立假說;zh-hk:備擇假設;zh-cn:备择假设
|3=zh-tw:虛無假說;zh-hk:虛無假設;zh-cn:零假设
|4=zh-cn:检验;zh-tw:檢定;
|5=zh-cn:水平;zh-tw:水準;
|6=zh-tw:假說;zh-hk:假設;zh-cn:假设
}}

[[File:Null and alternative distribution.jpg|thumb|一張包含虛無假說與對立假說兩個曲線的示意圖，兩[[常態分布]]有不同的[[期望值]]與相同的[[變異數]]]]

在[[推論統計學]]中，'''零假设'''（{{lang-en|Null hypothesis}}，又译'''-{zh-cn:虚无假设、原假设; zh-tw:零假說;}-'''，符号：<math>H_0</math>）是做[[假說檢定|统计检验]]时的一类[[假說]]。

零假设的内容一般是希望能证明为错误的假设，与零假设相对的是[[對立假說]]，即希望通过证伪零假设而证明正确的另一种假说。从数学上来看，零假设和备择假设的地位是相等的，但是在统计学的实际运用中，常常需要强调一类假设为应当或期望实现的假设，例如在相关性检验中，一般会取“两者之间无关联”作为零假设，而在[[皮爾森卡方檢定|独立性检验]]中，一般会取“两者之间非[[独立 (概率论)|獨立]]”作为零假设。

如果一个统计检验的结果'''拒绝'''（reject）零假设（结论不支持零假设），而实际上真实的情况属于零假设，那么称这个检验犯了[[第一型及第二型錯誤|型一錯誤]]。反之，如果检验结果支持零假设，而实际上真实的情况属于备择假设，那么称这个检验犯了[[型一錯誤與型二錯誤|型二錯誤]]。通常的做法是，在保持第一类错误出现的机会在某个特定水平上的时候（即[[显著性差异|显著性差异值]]或α值），尽量减少第二类错误出现的概率。

==相关条目==
*[[假說檢定]]
*[[對立假說]]
*[[檢定力]]
*[[型一錯誤與型二錯誤]]
*[[顯著性差異]]
*{{link-en|似然比检验|Likelihood-ratio test}}
*[[司徒頓t檢定]]

==参考来源==
*[https://web.archive.org/web/20100109044021/http://210.28.216.200/cai/tongji/zhishidian/knowledge5.htm 统计学原理：假设检验]

==延伸閱讀==
{{refbegin|2}}
*{{Cite book |last=Adèr |first=H. J. | author-link1 = Herman J. Adèr|last2= Mellenbergh  |first2= G. J.| author2-link = Gideon J. Mellenbergh |lastauthoramp=yes |last3=Hand |first3=D. J. |year=2007 |title=Advising on research methods: A consultant's companion |publisher=Johannes van Kessel Publishing |location=Huizen, The Netherlands |isbn=90-79418-01-3 }}
*{{Cite journal | doi = 10.1198/016214504000000089| title = Large-Scale Simultaneous Hypothesis Testing| url = https://archive.org/details/sim_journal-of-the-american-statistical-association_2004-03_99_465/page/96| journal = Journal of the American Statistical Association| volume = 99| issue = 465| pages = 96| year = 2004| last1 = Efron | first1 = B. }}  The application of significance testing in this paper is an outlier.  Tests to find a null hypothesis?  Not trying to show significance, but to find interesting cases?
*{{cite journal | last1 = Rice | first1 = William R. | last2 = Gaines | first2 = Steven D.  | title = 'Heads I win, tails you lose': testing directional alternative hypotheses in ecological and evolutionary research | journal = TREE | volume = 9 | number = 6 | pages = 235–237 | date = June 1994 | doi=10.1016/0169-5347(94)90258-5}}  Directed tests combine the attributes of one-tailed and two-tailed tests.  "...directed tests should be used in virtually all applications where one-sided tests have previously been used, excepting those cases where the data can only deviate from H<sub>0</sub>, in one direction."
{{refend}}

==外部連結==
* [http://davidmlane.com/hyperstat/A29337.html HyperStat Online: Null hypothesis] {{Wayback|url=http://davidmlane.com/hyperstat/A29337.html |date=20100210184835 }}

{{統計學}}

[[Category:统计检验]]
[[Category:统计学]]