t*****2
发帖数: 94 | 1 今天去面试一家公司，问了一些统计（基础）如GLM，power, type I error,
permutation test, logistic model 等问题，同时也问了优化的相关的问题。然后要
求上机解决一个实际问题,当场写出报告。
问题如下：
前面介绍了很多背景知识，简单来说如下(记得大致如下)
六个变量，共有600 observations
ID Age FEV Height Sex(M,F) Smoker(non=nonsmoker, current=current smoker)
301 9 1.708 57 Female Non |
A*******s
发帖数: 3942 | 2 how the residual plots look like? i guess it is critical from your
description...
smoker)
【在 t*****2 的大作中提到】
: 今天去面试一家公司，问了一些统计（基础）如GLM，power, type I error,
: permutation test, logistic model 等问题，同时也问了优化的相关的问题。然后要
: 求上机解决一个实际问题,当场写出报告。
: 问题如下：
: 前面介绍了很多背景知识，简单来说如下(记得大致如下)
: 六个变量，共有600 observations
: ID Age FEV Height Sex(M,F) Smoker(non=nonsmoker, current=current smoker)
: 301 9 1.708 57 Female Non
|
t*****2
发帖数: 94 | 3 我想还是做的步骤比较重要，表述都是基于步骤来说呀
如果你来做，你会做哪些呢？ |
A*******s
发帖数: 3942 | 4 case by case.
for your case generally i would say,
1. check Y's properties. is it bounded/truncated?
2. fit a linear model as benchmark
3. then u can talk one day about how to analyze residual plots
4. continue to talk one more day about how to analyze residual plots
5. continue to talk one more day about how to analyze residual plots
6. continue to talk one more day about how to analyze residual plots
7 ....
【在 t*****2 的大作中提到】
: 我想还是做的步骤比较重要，表述都是基于步骤来说呀
: 如果你来做，你会做哪些呢？
|
F****n
发帖数: 3271 | 5 Did you mentioned the potential non-linearity of age.
smoker)
【在 t*****2 的大作中提到】
: 今天去面试一家公司，问了一些统计（基础）如GLM，power, type I error,
: permutation test, logistic model 等问题，同时也问了优化的相关的问题。然后要
: 求上机解决一个实际问题,当场写出报告。
: 问题如下：
: 前面介绍了很多背景知识，简单来说如下(记得大致如下)
: 六个变量，共有600 observations
: ID Age FEV Height Sex(M,F) Smoker(non=nonsmoker, current=current smoker)
: 301 9 1.708 57 Female Non
|
t*****2
发帖数: 94 | |
s**f
发帖数: 365 | 7 对于continuous的variable，都加上至少quadratic和cubic的term，然后test是不是
significant
【在 t*****2 的大作中提到】
: 没有。对于AGE该怎么处理呢？
|
N****n
发帖数: 1208 | 8 为什么要这样？ 经验？
【在 s**f 的大作中提到】
: 对于continuous的variable，都加上至少quadratic和cubic的term，然后test是不是
: significant
|
F****n
发帖数: 3271 | 9 Discretize and / or test quadratic term. It's technically simple but kind of testing experience. The non-linearity of age should be always be addressed, especially this is about smoking.
For example, think about children / adult. Even the data are all adults, non-linear effects can also occur.
【在 t*****2 的大作中提到】
: 没有。对于AGE该怎么处理呢？
|
p*********o
发帖数: 138 | |
l****u
发帖数: 529 | 11 During child segment, age and height are kind of linearly related. So in
this period, multicollinearity should be pointed out.
kind of testing experience. The non-linearity of age should be always be
addressed, especially this is about smoking.
non-linear effects can also occur.
【在 F****n 的大作中提到】
: Discretize and / or test quadratic term. It's technically simple but kind of testing experience. The non-linearity of age should be always be addressed, especially this is about smoking.
: For example, think about children / adult. Even the data are all adults, non-linear effects can also occur.
|
t*****2
发帖数: 94 | |
i*****y
发帖数: 126 | |
j*******2
发帖数: 309 | 14 also check for outliers |
