d****y
发帖数: 58 | 1 刚面完。答的不咋样。
1. suppose four colors are there, we did a survey. People (n) choose
different colors (red, green, yellow, blue).
we get the proportion of people who use red as x1/n.
What is the accuracy for this?
2. Imaging now that we know from prior studies that a certain proportion of
users responds to questions randomly. How can we assess this proportion of "
random clickers"? |
e***m
发帖数: 92 | 2 第一题用大数定理吧
of
"
【在 d****y 的大作中提到】
: 刚面完。答的不咋样。
: 1. suppose four colors are there, we did a survey. People (n) choose
: different colors (red, green, yellow, blue).
: we get the proportion of people who use red as x1/n.
: What is the accuracy for this?
: 2. Imaging now that we know from prior studies that a certain proportion of
: users responds to questions randomly. How can we assess this proportion of "
: random clickers"?
|
c********h
发帖数: 330 | |
d****y
发帖数: 58 | 4 咋用？
n可大可小。
他有提到bootstrap.
【在 e***m 的大作中提到】
: 第一题用大数定理吧
:
: of
: "
|
f*****y
发帖数: 822 | 5 What's the solution for the second question?
Any insight? |
j*******g
发帖数: 331 | |
d****y
发帖数: 58 | 7 应该都是sampling的问题。要不然都没有mean, variance，咋求confidence interval.
当时面的时候就觉得很奇怪。反正说默认有这些真实值去解了。 |
h****y
发帖数: 33 | 8 第一个用不到multinomial，直接binomial就可以吧。是红色或者不是红色。需要用到
bootstrap吗？直接binomial分布的方差不就可以了吗。假设p hat是正态分布，然后得
到95%的置信区间。
第二题跟第一题其实是一个意思吧。都是讲proportion estimate的。 |
f*****y
发帖数: 822 | 9 第一题可以直接简化为binomial吧。
第二题的assess是什么意思，就是已经知道了这个proportion的数值，还是说怎么找出
这些random click？ |
