y*****w
发帖数: 1350 | 1 Longitudinal analysis, repeated measures analysis, and time series analysis,
谁能给个权威的比较（出处）？它们的定义，相似处，不同处，各自的适用范围？谢
谢！ |
z**********i
发帖数: 12276 | 2 我以为他们是一个意思，惭愧。
analysis,
【在 y*****w 的大作中提到】
: Longitudinal analysis, repeated measures analysis, and time series analysis,
: 谁能给个权威的比较（出处）？它们的定义，相似处，不同处，各自的适用范围？谢
: 谢！
|
Y****a
发帖数: 243 | 3 先给个不是权威的比较，抛砖引玉
repeated measuer 很多情况下是用于experimental design，用来比较两个treatments
的影响，效果之类。实验个体相同，意在消除不同sample可能引入的无法衡量的
baseline差别。两个treatment之间会加入一个washout period，assumption经过这个
washout period之后，前一组treatment的影响会全部消失，实验个体全部回归初始状
态。
longitudinal 一般是observational的，在多个时间点观测同一组变量，并研究这组变
量随着时间的变化。
time series 是指一种统计模型/分析方法。时间序列里的观测点一般比longitudinal
的要密集一些，大多意于预测，希望被观测的变量的变化是与时间相关的，具有某种可
模型化的规律。
他们之间在某种情况下可互换使用，还是主要看应用的场合和学科。 |
g**u
发帖数: 205 | 4 washout period不是只有在cross over design的时候才会出现吗？
treatments
longitudinal
【在 Y****a 的大作中提到】
: 先给个不是权威的比较，抛砖引玉
: repeated measuer 很多情况下是用于experimental design，用来比较两个treatments
: 的影响，效果之类。实验个体相同，意在消除不同sample可能引入的无法衡量的
: baseline差别。两个treatment之间会加入一个washout period，assumption经过这个
: washout period之后，前一组treatment的影响会全部消失，实验个体全部回归初始状
: 态。
: longitudinal 一般是observational的，在多个时间点观测同一组变量，并研究这组变
: 量随着时间的变化。
: time series 是指一种统计模型/分析方法。时间序列里的观测点一般比longitudinal
: 的要密集一些，大多意于预测，希望被观测的变量的变化是与时间相关的，具有某种可
|
w*******9
发帖数: 1433 | 5 longitudinal data is a subset of repeatedly measured data for more than one
subject. The focus of analysis is usually centered on the mean pattern not
the correlation between measurements (which is considered an important
nuisance parameter); time series could be considered as another form of
repeatedly measured data, but only for 1 subject. The focus is centered on
the correlation between adjacent measurements. Since there's only one
subject, you
need a longer chain. Understanding this difference will help you to
understand what the models are trying to model. |
d***u
发帖数: 60 | 6 我怎么觉得你说的正好相反啊。。。。难道我记错了。。
one
【在 w*******9 的大作中提到】
: longitudinal data is a subset of repeatedly measured data for more than one
: subject. The focus of analysis is usually centered on the mean pattern not
: the correlation between measurements (which is considered an important
: nuisance parameter); time series could be considered as another form of
: repeatedly measured data, but only for 1 subject. The focus is centered on
: the correlation between adjacent measurements. Since there's only one
: subject, you
: need a longer chain. Understanding this difference will help you to
: understand what the models are trying to model.
|
s*****n
发帖数: 2174 | 7 +1
一般longitudinal都有多个subjects, 每个subject有若干observations over time.
time series一般只有一个subject, 但是这个subject有很多observations over time.
所以longitudinal study需要考虑样本间的统计性质, 只是要注意observations之间的
intra-correlation.
而time series study一般不是考虑样本间的统计性质, 而是注重在自身的这一个序列
内部找关系.
one
【在 w*******9 的大作中提到】
: longitudinal data is a subset of repeatedly measured data for more than one
: subject. The focus of analysis is usually centered on the mean pattern not
: the correlation between measurements (which is considered an important
: nuisance parameter); time series could be considered as another form of
: repeatedly measured data, but only for 1 subject. The focus is centered on
: the correlation between adjacent measurements. Since there's only one
: subject, you
: need a longer chain. Understanding this difference will help you to
: understand what the models are trying to model.
|
