Logistic回归分析中,连续型变量OR值太大,是什么原因?

2019-08-22 不详 网络

Logistic回归是临床研究中最常见的多因素分析方法之一,但是,各类问题也十分多见的。其中有一个问题经常困扰,即分析时发现某个变量的OR值过大的问题。一般来说,大部分情况,一个因素对结局的影响都不可能过大(因为,都是各种因素的综合效应)。如果是超大样本分析中,OR值很少有超过2的。而在日常分析中,OR经常会达到10以上,甚至100以上,这是什么回事呢?看一个案例: 下面是几种处理方

Logistic回归是临床研究中最常见的多因素分析方法之一,但是,各类问题也十分多见的。其中有一个问题经常困扰,即分析时发现某个变量的OR值过大的问题。一般来说,大部分情况,一个因素对结局的影响都不可能过大(因为,都是各种因素的综合效应)。如果是超大样本分析中,OR值很少有超过2的。

而在日常分析中,OR经常会达到10以上,甚至100以上,这是什么回事呢?看一个案例:



下面是几种处理方法,你认为呢?

1、改一下X(连续性自变量)的单位即可,例如X是身高,原来单位是m,那么得出来的OR表示每增加1m的危险比。现在把它换成cm, 得出来的则是每增加1cm的危险比了。OR就小多了。

2、建议对连续型变量进行分段处理,转变为分类变量试试!

3、如果其结果和临床相差太大可能是样本收集问提或共线性问题。

4、更要防止数据本身的问题:

  • 该变量某一类的例数特别少,如性别,男性有100人,女性有2人,可能会出现这种情形。

  • 空单元格(zero cell count),如性别与疾病的关系,所有男性都发生了疾病或都没有发生疾病,这时候可能会出现OR值无穷大或为0的情形。

  • 完全分离(complete separation),对于某自变量,如果该自变量取值大于某一值时结局发生,当小于该值时结局都不发生,就会出现完全分离现象。如年龄20~、30~、40~、50~四个年龄段,如果40岁以上的人全部发生疾病,40岁以下的人全部不发病,就就产生了完全分离现象,也会出现一个大得不可理喻的标准误。

  • 多重共线性问题,多重共线性会产生大的标准误。

  • 样本量过小,变量过少,导致其中一个变量与结局的关联度过高。


版权声明:
本网站所有内容来源注明为“梅斯医学”或“MedSci原创”的文字、图片和音视频资料,版权均属于梅斯医学所有。非经授权,任何媒体、网站或个人不得转载,授权转载时须注明来源为“梅斯医学”。其它来源的文章系转载文章,或“梅斯号”自媒体发布的文章,仅系出于传递更多信息之目的,本站仅负责审核内容合规,其内容不代表本站立场,本站不负责内容的准确性和版权。如果存在侵权、或不希望被转载的媒体或个人可与我们联系,我们将立即进行删除处理。
在此留言
评论区 (4)
#插入话题
  1. [GetPortalCommentsPageByObjectIdResponse(id=1746740, encodeId=9a371e46740e0, content=<a href='/topic/show?id=88e91099e10' target=_blank style='color:#2F92EE;'>#logistic回归#</a>, beContent=null, objectType=article, channel=null, level=null, likeNumber=45, replyNumber=0, topicName=null, topicId=null, topicList=[TopicDto(id=10997, encryptionId=88e91099e10, topicName=logistic回归)], attachment=null, authenticateStatus=null, createdAvatar=null, createdBy=d3c135838005, createdName=xlwang2703, createdTime=Thu Feb 13 14:23:00 CST 2020, time=2020-02-13, status=1, ipAttribution=), GetPortalCommentsPageByObjectIdResponse(id=1671703, encodeId=70f716e1703ef, content=<a href='/topic/show?id=c56c4051902' target=_blank style='color:#2F92EE;'>#回归分析#</a>, beContent=null, objectType=article, channel=null, level=null, likeNumber=51, replyNumber=0, topicName=null, topicId=null, topicList=[TopicDto(id=40519, encryptionId=c56c4051902, topicName=回归分析)], attachment=null, authenticateStatus=null, createdAvatar=null, createdBy=a46f26833520, createdName=cenghis, createdTime=Thu Jul 23 14:23:00 CST 2020, time=2020-07-23, status=1, ipAttribution=), GetPortalCommentsPageByObjectIdResponse(id=1905770, encodeId=cfa21905e70a1, content=<a href='/topic/show?id=9dc4109964f' target=_blank style='color:#2F92EE;'>#Logistic#</a>, beContent=null, objectType=article, channel=null, level=null, likeNumber=44, replyNumber=0, topicName=null, topicId=null, topicList=[TopicDto(id=10996, encryptionId=9dc4109964f, topicName=Logistic)], attachment=null, authenticateStatus=null, createdAvatar=, createdBy=afa4194, createdName=cenghis, createdTime=Sat Jul 18 11:23:00 CST 2020, time=2020-07-18, status=1, ipAttribution=), GetPortalCommentsPageByObjectIdResponse(id=1455154, encodeId=aa3c1455154eb, content=<a href='/topic/show?id=bebe8032fc' target=_blank style='color:#2F92EE;'>#GIST#</a>, beContent=null, objectType=article, channel=null, level=null, likeNumber=36, replyNumber=0, topicName=null, topicId=null, topicList=[TopicDto(id=8032, encryptionId=bebe8032fc, topicName=GIST)], attachment=null, authenticateStatus=null, createdAvatar=null, createdBy=6e445632366, createdName=july_977, createdTime=Sat Aug 24 00:23:00 CST 2019, time=2019-08-24, status=1, ipAttribution=)]
  2. [GetPortalCommentsPageByObjectIdResponse(id=1746740, encodeId=9a371e46740e0, content=<a href='/topic/show?id=88e91099e10' target=_blank style='color:#2F92EE;'>#logistic回归#</a>, beContent=null, objectType=article, channel=null, level=null, likeNumber=45, replyNumber=0, topicName=null, topicId=null, topicList=[TopicDto(id=10997, encryptionId=88e91099e10, topicName=logistic回归)], attachment=null, authenticateStatus=null, createdAvatar=null, createdBy=d3c135838005, createdName=xlwang2703, createdTime=Thu Feb 13 14:23:00 CST 2020, time=2020-02-13, status=1, ipAttribution=), GetPortalCommentsPageByObjectIdResponse(id=1671703, encodeId=70f716e1703ef, content=<a href='/topic/show?id=c56c4051902' target=_blank style='color:#2F92EE;'>#回归分析#</a>, beContent=null, objectType=article, channel=null, level=null, likeNumber=51, replyNumber=0, topicName=null, topicId=null, topicList=[TopicDto(id=40519, encryptionId=c56c4051902, topicName=回归分析)], attachment=null, authenticateStatus=null, createdAvatar=null, createdBy=a46f26833520, createdName=cenghis, createdTime=Thu Jul 23 14:23:00 CST 2020, time=2020-07-23, status=1, ipAttribution=), GetPortalCommentsPageByObjectIdResponse(id=1905770, encodeId=cfa21905e70a1, content=<a href='/topic/show?id=9dc4109964f' target=_blank style='color:#2F92EE;'>#Logistic#</a>, beContent=null, objectType=article, channel=null, level=null, likeNumber=44, replyNumber=0, topicName=null, topicId=null, topicList=[TopicDto(id=10996, encryptionId=9dc4109964f, topicName=Logistic)], attachment=null, authenticateStatus=null, createdAvatar=, createdBy=afa4194, createdName=cenghis, createdTime=Sat Jul 18 11:23:00 CST 2020, time=2020-07-18, status=1, ipAttribution=), GetPortalCommentsPageByObjectIdResponse(id=1455154, encodeId=aa3c1455154eb, content=<a href='/topic/show?id=bebe8032fc' target=_blank style='color:#2F92EE;'>#GIST#</a>, beContent=null, objectType=article, channel=null, level=null, likeNumber=36, replyNumber=0, topicName=null, topicId=null, topicList=[TopicDto(id=8032, encryptionId=bebe8032fc, topicName=GIST)], attachment=null, authenticateStatus=null, createdAvatar=null, createdBy=6e445632366, createdName=july_977, createdTime=Sat Aug 24 00:23:00 CST 2019, time=2019-08-24, status=1, ipAttribution=)]
  3. [GetPortalCommentsPageByObjectIdResponse(id=1746740, encodeId=9a371e46740e0, content=<a href='/topic/show?id=88e91099e10' target=_blank style='color:#2F92EE;'>#logistic回归#</a>, beContent=null, objectType=article, channel=null, level=null, likeNumber=45, replyNumber=0, topicName=null, topicId=null, topicList=[TopicDto(id=10997, encryptionId=88e91099e10, topicName=logistic回归)], attachment=null, authenticateStatus=null, createdAvatar=null, createdBy=d3c135838005, createdName=xlwang2703, createdTime=Thu Feb 13 14:23:00 CST 2020, time=2020-02-13, status=1, ipAttribution=), GetPortalCommentsPageByObjectIdResponse(id=1671703, encodeId=70f716e1703ef, content=<a href='/topic/show?id=c56c4051902' target=_blank style='color:#2F92EE;'>#回归分析#</a>, beContent=null, objectType=article, channel=null, level=null, likeNumber=51, replyNumber=0, topicName=null, topicId=null, topicList=[TopicDto(id=40519, encryptionId=c56c4051902, topicName=回归分析)], attachment=null, authenticateStatus=null, createdAvatar=null, createdBy=a46f26833520, createdName=cenghis, createdTime=Thu Jul 23 14:23:00 CST 2020, time=2020-07-23, status=1, ipAttribution=), GetPortalCommentsPageByObjectIdResponse(id=1905770, encodeId=cfa21905e70a1, content=<a href='/topic/show?id=9dc4109964f' target=_blank style='color:#2F92EE;'>#Logistic#</a>, beContent=null, objectType=article, channel=null, level=null, likeNumber=44, replyNumber=0, topicName=null, topicId=null, topicList=[TopicDto(id=10996, encryptionId=9dc4109964f, topicName=Logistic)], attachment=null, authenticateStatus=null, createdAvatar=, createdBy=afa4194, createdName=cenghis, createdTime=Sat Jul 18 11:23:00 CST 2020, time=2020-07-18, status=1, ipAttribution=), GetPortalCommentsPageByObjectIdResponse(id=1455154, encodeId=aa3c1455154eb, content=<a href='/topic/show?id=bebe8032fc' target=_blank style='color:#2F92EE;'>#GIST#</a>, beContent=null, objectType=article, channel=null, level=null, likeNumber=36, replyNumber=0, topicName=null, topicId=null, topicList=[TopicDto(id=8032, encryptionId=bebe8032fc, topicName=GIST)], attachment=null, authenticateStatus=null, createdAvatar=null, createdBy=6e445632366, createdName=july_977, createdTime=Sat Aug 24 00:23:00 CST 2019, time=2019-08-24, status=1, ipAttribution=)]
    2020-07-18 cenghis
  4. [GetPortalCommentsPageByObjectIdResponse(id=1746740, encodeId=9a371e46740e0, content=<a href='/topic/show?id=88e91099e10' target=_blank style='color:#2F92EE;'>#logistic回归#</a>, beContent=null, objectType=article, channel=null, level=null, likeNumber=45, replyNumber=0, topicName=null, topicId=null, topicList=[TopicDto(id=10997, encryptionId=88e91099e10, topicName=logistic回归)], attachment=null, authenticateStatus=null, createdAvatar=null, createdBy=d3c135838005, createdName=xlwang2703, createdTime=Thu Feb 13 14:23:00 CST 2020, time=2020-02-13, status=1, ipAttribution=), GetPortalCommentsPageByObjectIdResponse(id=1671703, encodeId=70f716e1703ef, content=<a href='/topic/show?id=c56c4051902' target=_blank style='color:#2F92EE;'>#回归分析#</a>, beContent=null, objectType=article, channel=null, level=null, likeNumber=51, replyNumber=0, topicName=null, topicId=null, topicList=[TopicDto(id=40519, encryptionId=c56c4051902, topicName=回归分析)], attachment=null, authenticateStatus=null, createdAvatar=null, createdBy=a46f26833520, createdName=cenghis, createdTime=Thu Jul 23 14:23:00 CST 2020, time=2020-07-23, status=1, ipAttribution=), GetPortalCommentsPageByObjectIdResponse(id=1905770, encodeId=cfa21905e70a1, content=<a href='/topic/show?id=9dc4109964f' target=_blank style='color:#2F92EE;'>#Logistic#</a>, beContent=null, objectType=article, channel=null, level=null, likeNumber=44, replyNumber=0, topicName=null, topicId=null, topicList=[TopicDto(id=10996, encryptionId=9dc4109964f, topicName=Logistic)], attachment=null, authenticateStatus=null, createdAvatar=, createdBy=afa4194, createdName=cenghis, createdTime=Sat Jul 18 11:23:00 CST 2020, time=2020-07-18, status=1, ipAttribution=), GetPortalCommentsPageByObjectIdResponse(id=1455154, encodeId=aa3c1455154eb, content=<a href='/topic/show?id=bebe8032fc' target=_blank style='color:#2F92EE;'>#GIST#</a>, beContent=null, objectType=article, channel=null, level=null, likeNumber=36, replyNumber=0, topicName=null, topicId=null, topicList=[TopicDto(id=8032, encryptionId=bebe8032fc, topicName=GIST)], attachment=null, authenticateStatus=null, createdAvatar=null, createdBy=6e445632366, createdName=july_977, createdTime=Sat Aug 24 00:23:00 CST 2019, time=2019-08-24, status=1, ipAttribution=)]
    2019-08-24 july_977

相关资讯

Logistic回归(1)

本章课程围绕Logistic回归展开,一共分为四小节。第一节课程,通过一幅图形象生动的了解Logistic回归的本质及思想。

Logistic回归(2)

本节课程讲解Logistic回归模型的建立,以及一些重要的细节问题,包括多重共线性、变量筛选方法、交互作用、哑变量设置、样本量等。

BMJ:近期感染降低类风湿关节炎患病风险

近期感染影响类风湿关节炎(RA)患病风险吗?读完这篇研究告诉你答案。研究者使用基于人群的个体控制研究EIRA(n=6401),以观察和健康对照组相比,RA的发病风险。该研究对研究对象按照性别、年龄、时间、居住地区域进行匹配。入组前的两年内有无发生胃肠炎、尿路感染、生殖器感染、前列腺炎、鼻窦炎、扁桃体炎、肺炎的感染情况被统计记录,使用假设Logistic 回归分析来计算OR值,吸烟情况和社会经济地位

二分类logistic回归中纳入多少自变量合适?

在对临床数据的探索分析工作中,我们经常会使用多因素logistic回归分析去探索疾病的危险因素,也可以用它来做预测。但是每每在使用logistic回归分析的时候,我们都会纠结应该选哪些作为自变量呢?选多少个合适呢? 选哪些作为自变量,这个问题比较简单。一般情况下,我都是选择那些单因素分析中与因变量有关的自变量进入回归方程。但同时需要提醒,如果某些自变量从作用机制或临床经验上来看,

Logistic回归(3)

本节课程讲解Logistic回归中的优势、RR/OR/HR的含义,以及在医学中的应用。

似然比与logistic回归

似然,英文就是likelihood,实际上也可以翻译为可能性,翻译成“似然”可能只是为了书面听起来好听一些。最大似然也就是最大可能性,最大似然估计也就是找出一个最大可能使函数接近观测值的一个参数。通 常我们用统计软件做卡方检验时,同时会有一个似然比检验,而且二者的值极其接近。卡方检验和似然比检验其实并没有什么太大关系,更像是两条可以同时到达同 一个终点的路,至于你喜欢走哪条路,你自己选择,并没