c****o 发帖数: 69 | 1 貌似是好几个月前头的,刚收到邮件,要约面试,我在银行工作了3年,对IT行业的
STATISTICIAN还真的是一点概念都没有,请问有了解或者面过他家的同学给点提示么?
?需要考DETAILED的编程么。。。还是只是聊PROJECT?
PS:我对编程的了解,只限于SAS,SQL, 不知道他家会不会问很多C++或者MATLAB的东西
呢?? |
C******y 发帖数: 2007 | |
c****o 发帖数: 69 | 3 thanks so much.
I actually read those posts before and looks like I'm far away from being
able to answer the questions....lol
【在 C******y 的大作中提到】![](/moin_static193/solenoid/img/up.png) : 本版有google面经
|
w**********y 发帖数: 1691 | 4 听说过的问题:
你的research必问.
open question一步步的引导你去回答一个跟google的business相关的model问题.忽然
想不起来了....
让你估计西雅图的大楼一共有多少玻璃
我认识个那个统计phd,c++一点也没问.问了一些R |
B******5 发帖数: 4676 | |
p**n 发帖数: 1437 | |
m**********4 发帖数: 774 | 7 我们系两个拿到google的人都不会c++或者 sql
其中一个人面食的时候我在旁边。一堆r的问题,还记得一个glm的问题。
别的就没啥了,我觉得不是很困难。
当然我那两个同学都是research大牛。
【在 c****o 的大作中提到】![](/moin_static193/solenoid/img/up.png) : 貌似是好几个月前头的,刚收到邮件,要约面试,我在银行工作了3年,对IT行业的 : STATISTICIAN还真的是一点概念都没有,请问有了解或者面过他家的同学给点提示么? : ?需要考DETAILED的编程么。。。还是只是聊PROJECT? : PS:我对编程的了解,只限于SAS,SQL, 不知道他家会不会问很多C++或者MATLAB的东西 : 呢??
|
B******5 发帖数: 4676 | 8 赞牛系牛人们
【在 m**********4 的大作中提到】![](/moin_static193/solenoid/img/up.png) : 我们系两个拿到google的人都不会c++或者 sql : 其中一个人面食的时候我在旁边。一堆r的问题,还记得一个glm的问题。 : 别的就没啥了,我觉得不是很困难。 : 当然我那两个同学都是research大牛。
|
c****o 发帖数: 69 | 9 Update on interview:
1st phone interview:
1. question mostly related to my working experience
2. how to generate random number in R (havent been using R for a
while....and totally forget it...)
3. Senior high school students go to see college admission committee(10
members totally). Each committee member will give a result for the
candidate: accepted/rejected/waiting list. 1 year later, we would have
how those students doing at college. Question: how can we determine
whether the committee was doing a good job or not/or how can we
determine whether each member was doing a good job or not
1st phone interview went very smoothly (except R question), but the 2nd
one was a disaster for me but it was not hard actually. Just becoz I
havent been doing those calculation for years after school.....anyway,
hopefully it helps for other ppl:
2nd phone interview:
1. Questions related to one of the logistic regression model I built:
how to select variable, how to do validation, how to use it, how to
improve it;
2. there are 5 toys at KFC, and if you go there, u will get one toy
totally randomly. How many times you need to go to KFC in order to get
all 5 toys ---- I died here.....lol......
3. for binary response variable, if we build the model using linear
regression model, what would be the problem, what's the assumption of
linear regression model would then break; how to solve it (using
logistic regression model for sure....); Why using Logistic regression
model would solve this problem???
Good luck for other ppl who will be interviewing with Google. |
k*******a 发帖数: 772 | 10 KFC那个很简单啊
1+1/0.8+1/0.6+1/0.4+1/0.2
【在 c****o 的大作中提到】![](/moin_static193/solenoid/img/up.png) : Update on interview: : 1st phone interview: : 1. question mostly related to my working experience : 2. how to generate random number in R (havent been using R for a : while....and totally forget it...) : 3. Senior high school students go to see college admission committee(10 : members totally). Each committee member will give a result for the : candidate: accepted/rejected/waiting list. 1 year later, we would have : how those students doing at college. Question: how can we determine : whether the committee was doing a good job or not/or how can we
|
|
|
h******s 发帖数: 3420 | 11 their questions are not hard, but they expect a perfect answer. I know
quite a few people around me got rejected for very minor mistakes (a typo
in the code?)
【在 c****o 的大作中提到】![](/moin_static193/solenoid/img/up.png) : 貌似是好几个月前头的,刚收到邮件,要约面试,我在银行工作了3年,对IT行业的 : STATISTICIAN还真的是一点概念都没有,请问有了解或者面过他家的同学给点提示么? : ?需要考DETAILED的编程么。。。还是只是聊PROJECT? : PS:我对编程的了解,只限于SAS,SQL, 不知道他家会不会问很多C++或者MATLAB的东西 : 呢??
|
m****e 发帖数: 255 | 12 Allow me shred some light on who/how-to be hired as S*T*A*T*I*S*T*I*C*I*A*N at G$O$O
$G$L$E.
1. Get your master or PhD in top five (maybe top 10) stats program or have a
well known PhD advisor
2. Be local or at least in California.
3. Important to have intern/industry experience.
4. Be referred and have onsite directly.
If you have all of above, you will probably be hired. If you only get one or
two out of the four items, you better to be liked at first sight by your
interviewers (at least by some of them).
Cheers |
h******s 发帖数: 3420 | 13 统计也搞海选吗?CS 大概200个选一个。一帮子人啥也不干就光在那面试,典型吃饱了
撑的。别的公司都没这么恶心。至少apple 好很多。
N at G$O$O
a
or
【在 m****e 的大作中提到】![](/moin_static193/solenoid/img/up.png) : Allow me shred some light on who/how-to be hired as S*T*A*T*I*S*T*I*C*I*A*N at G$O$O : $G$L$E. : 1. Get your master or PhD in top five (maybe top 10) stats program or have a : well known PhD advisor : 2. Be local or at least in California. : 3. Important to have intern/industry experience. : 4. Be referred and have onsite directly. : If you have all of above, you will probably be hired. If you only get one or : two out of the four items, you better to be liked at first sight by your : interviewers (at least by some of them).
|
c****o 发帖数: 69 | 14 can you also share a lit bit about the compensation about statistician at
Google?
and what is the career path for Statistician role there?
N at G$O$O
a
or
【在 m****e 的大作中提到】![](/moin_static193/solenoid/img/up.png) : Allow me shred some light on who/how-to be hired as S*T*A*T*I*S*T*I*C*I*A*N at G$O$O : $G$L$E. : 1. Get your master or PhD in top five (maybe top 10) stats program or have a : well known PhD advisor : 2. Be local or at least in California. : 3. Important to have intern/industry experience. : 4. Be referred and have onsite directly. : If you have all of above, you will probably be hired. If you only get one or : two out of the four items, you better to be liked at first sight by your : interviewers (at least by some of them).
|
c****o 发帖数: 69 | 15 i dont think it would be hard but i just can not get the answer during the
interview...and actually i still didnt get it with the answer you provide...
....lol....i guess Im just not qualified for Google :)
【在 k*******a 的大作中提到】![](/moin_static193/solenoid/img/up.png) : KFC那个很简单啊 : 1+1/0.8+1/0.6+1/0.4+1/0.2
|
g******n 发帖数: 339 | 16 这个不太明白,如果每次都是独立的事件, 一直拿不到所有的玩具的概率总是大于0.
所以说,只能说是至少多少次以后,能够拿到所有的玩具的概率是多少,比如95%。用
如下简单R-code的simulation,基于multinomial distrubition,20次以上可以保证拿
到所有玩具的概率至少是94%。不过要让我推公式算,要是面试肯定没戏。
Prob<-function(N=10,nsim=1000,level=0.9,seed=1324){
set.seed(seed)
x<-rmultinom(nsim, size=N, c(0.2,0.2,0.2,0.2,0.2))
y<-numeric(length=nsim)+1
for(i in 1:nsim){
if (any(x[,i]==0))
y[i]<-0
}
return(sum(y)/length(y))
}
【在 k*******a 的大作中提到】![](/moin_static193/solenoid/img/up.png) : KFC那个很简单啊 : 1+1/0.8+1/0.6+1/0.4+1/0.2
|
h**********1 发帖数: 155 | 17 i agree with you.
but this is the average number of visits in order to get all 5 toys.
a little explanation: you can view each one has a geometric distribution and
sum up.
【在 k*******a 的大作中提到】![](/moin_static193/solenoid/img/up.png) : KFC那个很简单啊 : 1+1/0.8+1/0.6+1/0.4+1/0.2
|
k****n 发帖数: 165 | 18 NVM. mistake
【在 k*******a 的大作中提到】![](/moin_static193/solenoid/img/up.png) : KFC那个很简单啊 : 1+1/0.8+1/0.6+1/0.4+1/0.2
|
n****e 发帖数: 226 | 19 这个很像骰子的一个问题,平均多少次能把1到6都掷到
我用R做一个简单的 simulation
大概是 11.42,跟前面的一个回复很相近了
【在 g******n 的大作中提到】![](/moin_static193/solenoid/img/up.png) : 这个不太明白,如果每次都是独立的事件, 一直拿不到所有的玩具的概率总是大于0. : 所以说,只能说是至少多少次以后,能够拿到所有的玩具的概率是多少,比如95%。用 : 如下简单R-code的simulation,基于multinomial distrubition,20次以上可以保证拿 : 到所有玩具的概率至少是94%。不过要让我推公式算,要是面试肯定没戏。 : Prob<-function(N=10,nsim=1000,level=0.9,seed=1324){ : set.seed(seed) : : x<-rmultinom(nsim, size=N, c(0.2,0.2,0.2,0.2,0.2)) : y<-numeric(length=nsim)+1 : for(i in 1:nsim){
|
w********m 发帖数: 1137 | 20 Best case: get unique toy each time --> only need 5 visits
Worst case: each visit is independent --> to get a toy needs 5 times -->
to get 5 toys needs 25 times
Result: visits to get all 5 toys is [5, 25] |
|
|
m****e 发帖数: 255 | 21 In recent years, fresh PhD gets about 120k-130k, 10-15% target bonus (in the
end probably double that) and a number of stocks, not mentioning the
benefits. You can also check h1b wage for base salary.
Career path wise, depends on what you want then. The best strategy is to
jump to the next big thing.On the other hand, it is not easy to get to a
managerial position because team structure is flat.
If you haven't been officially rejected yet, you may have still a tiny bit chance
depending on who was the second interviewer.
【在 c****o 的大作中提到】![](/moin_static193/solenoid/img/up.png) : can you also share a lit bit about the compensation about statistician at : Google? : and what is the career path for Statistician role there? : : N at G$O$O : a : or
|
s*******n 发帖数: 631 | 22
the
chance
How about masters?Not needed at all?
【在 m****e 的大作中提到】![](/moin_static193/solenoid/img/up.png) : In recent years, fresh PhD gets about 120k-130k, 10-15% target bonus (in the : end probably double that) and a number of stocks, not mentioning the : benefits. You can also check h1b wage for base salary. : Career path wise, depends on what you want then. The best strategy is to : jump to the next big thing.On the other hand, it is not easy to get to a : managerial position because team structure is flat. : If you haven't been officially rejected yet, you may have still a tiny bit chance : depending on who was the second interviewer.
|
o****t 发帖数: 515 | 23 you can get the toy1 25 times even the chance is small.
【在 w********m 的大作中提到】![](/moin_static193/solenoid/img/up.png) : Best case: get unique toy each time --> only need 5 visits : Worst case: each visit is independent --> to get a toy needs 5 times --> : to get 5 toys needs 25 times : Result: visits to get all 5 toys is [5, 25]
|
c****o 发帖数: 69 | 24 not yet officially rejected tho I dont think i would have a chance.
Actually i would need to make decision on another offer in banking this week
. I guess i just have no chance for Google this time....
the
chance
【在 m****e 的大作中提到】![](/moin_static193/solenoid/img/up.png) : In recent years, fresh PhD gets about 120k-130k, 10-15% target bonus (in the : end probably double that) and a number of stocks, not mentioning the : benefits. You can also check h1b wage for base salary. : Career path wise, depends on what you want then. The best strategy is to : jump to the next big thing.On the other hand, it is not easy to get to a : managerial position because team structure is flat. : If you haven't been officially rejected yet, you may have still a tiny bit chance : depending on who was the second interviewer.
|
c****o 发帖数: 69 | 25 they hire master for sure.
【在 s*******n 的大作中提到】![](/moin_static193/solenoid/img/up.png) : : the : chance : How about masters?Not needed at all?
|
i***o 发帖数: 778 | 26 The idea is: Given the kinds of toys you have on hand, what is the
expectation rounds that you need to get a different kind of toy from what
you have now.
..
【在 c****o 的大作中提到】![](/moin_static193/solenoid/img/up.png) : i dont think it would be hard but i just can not get the answer during the : interview...and actually i still didnt get it with the answer you provide... : ....lol....i guess Im just not qualified for Google :)
|
e****e 发帖数: 2740 | 27 rt
【在 c****o 的大作中提到】![](/moin_static193/solenoid/img/up.png) : 貌似是好几个月前头的,刚收到邮件,要约面试,我在银行工作了3年,对IT行业的 : STATISTICIAN还真的是一点概念都没有,请问有了解或者面过他家的同学给点提示么? : ?需要考DETAILED的编程么。。。还是只是聊PROJECT? : PS:我对编程的了解,只限于SAS,SQL, 不知道他家会不会问很多C++或者MATLAB的东西 : 呢??
|
w********m 发帖数: 1137 | 28 The probability that Visiting Frequencies = 25 is 0.8^25 = 0.0038, so the
right tail beyond 25 than it can be ignored. The distribution of selecting 5
toys is skew to the right with a cutting value at 5.
Given limited time, my guess of the expected value is (5+25) / 2 = 15.
【在 o****t 的大作中提到】![](/moin_static193/solenoid/img/up.png) : you can get the toy1 25 times even the chance is small.
|
A*******s 发帖数: 3942 | 29 i think u r correct.
the total # of visits are just the sum of 5 independent geometric
distributed random variables, we can have expectation as well as variance as
the sum of each one's. Do interviews also want the exact form of
distribution? that would be quite difficult...at least for me.
and
【在 h**********1 的大作中提到】![](/moin_static193/solenoid/img/up.png) : i agree with you. : but this is the average number of visits in order to get all 5 toys. : a little explanation: you can view each one has a geometric distribution and : sum up.
|
c****o 发帖数: 69 | 30 he didnt ask me the exact form but just a simple question: how many times
you will need to go to KFC to get all the 5 toys....
as
【在 A*******s 的大作中提到】![](/moin_static193/solenoid/img/up.png) : i think u r correct. : the total # of visits are just the sum of 5 independent geometric : distributed random variables, we can have expectation as well as variance as : the sum of each one's. Do interviews also want the exact form of : distribution? that would be quite difficult...at least for me. : : and
|
|
|
A*******s 发帖数: 3942 | 31 i think giving him the expectation and variance is more than enough ba...
u cannot say how many times exactly since its random.
【在 c****o 的大作中提到】![](/moin_static193/solenoid/img/up.png) : he didnt ask me the exact form but just a simple question: how many times : you will need to go to KFC to get all the 5 toys.... : : as
|
D*********2 发帖数: 535 | 32 my 2 cents about KFC,
I guess lz might miss one piece of info about the confidence level, the
problem I am shooting for is "have at least a 95% chance of getting all the
toys".
Prob{ get all 5 toys } = 1 - 5 * ( 4/5 )^n + 10 * ( 3/5 )^n - 10 * ( 2/5 )^n
+ 5 * ( 1/5 )^n
plz do not ask me why, check ur math stat book first chapter. classic
problem.
one can also access the probability thru simulation, which is 老牛 did :-)
###################################################
SuccProb<-function(nrun=10,nsim=10000,seed=1324){
set.seed(seed)
x <- rmultinom(nsim, size=nrun, rep(1/5, 5))
y <- 1 - mean(apply(x==0, 2, any))
z <- 1 - 5*(.8)^nrun + 10*(.6)^nrun - 10*(.4)^nrun + 5*(.2)^nrun
return(round(c(y, z), 4))
}
for( i in 10:30 ) print(SuccProb(nrun=i))
###################################################
simu theory
10 0.5202 0.5225
11 0.5982 0.6064
12 0.6727 0.678
13 0.7308 0.7381
14 0.7874 0.7879
15 0.824 0.8288
16 0.8559 0.8621
17 0.8875 0.8891
18 0.9086 0.9109
19 0.9277 0.9286
20 0.9435 0.9427
21 0.9549 0.9541
22 0.9631 0.9632
23 0.9712 0.9706
24 0.9773 0.9764
25 0.9813 0.9811
26 0.9839 0.9849
27 0.9874 0.9879
28 0.9912 0.9903
29 0.9925 0.9923
30 0.9946 0.9938 |
A*******s 发帖数: 3942 | 33 cool... so there is an exact form.. need to review text :)
the
^n
【在 D*********2 的大作中提到】![](/moin_static193/solenoid/img/up.png) : my 2 cents about KFC, : I guess lz might miss one piece of info about the confidence level, the : problem I am shooting for is "have at least a 95% chance of getting all the : toys". : Prob{ get all 5 toys } = 1 - 5 * ( 4/5 )^n + 10 * ( 3/5 )^n - 10 * ( 2/5 )^n : + 5 * ( 1/5 )^n : plz do not ask me why, check ur math stat book first chapter. classic : problem. : one can also access the probability thru simulation, which is 老牛 did :-) : ###################################################
|
c****o 发帖数: 69 | 34 is the following way also calculating the rite answer:
P(get all 5 toys in N times) = P(N,5)*5^(N-5)/5^N?
the
^n
【在 D*********2 的大作中提到】![](/moin_static193/solenoid/img/up.png) : my 2 cents about KFC, : I guess lz might miss one piece of info about the confidence level, the : problem I am shooting for is "have at least a 95% chance of getting all the : toys". : Prob{ get all 5 toys } = 1 - 5 * ( 4/5 )^n + 10 * ( 3/5 )^n - 10 * ( 2/5 )^n : + 5 * ( 1/5 )^n : plz do not ask me why, check ur math stat book first chapter. classic : problem. : one can also access the probability thru simulation, which is 老牛 did :-) : ###################################################
|
c***d 发帖数: 70 | 35 google: Coupon collector's problem |
c***d 发帖数: 70 | 36 n(1/1+1/2+...+1/n)=11.4166667 |
g******n 发帖数: 339 | 37 Enough is enough. icced's post should be the show stopper:
http://en.wikipedia.org/wiki/Coupon_collector's_problem
It seems that this is a classic problem and people often talk about the mean
number of trials. You(myself included) are trying to answer the question
from a different aspect, which I still think makes sense in a centain way.
the
^n
【在 D*********2 的大作中提到】![](/moin_static193/solenoid/img/up.png) : my 2 cents about KFC, : I guess lz might miss one piece of info about the confidence level, the : problem I am shooting for is "have at least a 95% chance of getting all the : toys". : Prob{ get all 5 toys } = 1 - 5 * ( 4/5 )^n + 10 * ( 3/5 )^n - 10 * ( 2/5 )^n : + 5 * ( 1/5 )^n : plz do not ask me why, check ur math stat book first chapter. classic : problem. : one can also access the probability thru simulation, which is 老牛 did :-) : ###################################################
|
D*********2 发帖数: 535 | 38 那啥,不管有没有人关心,我再多嘴一句啊~
老牛跟我算的是CDF,题目问的是Expectation。要是有人想验证的话,算
sum( ( F[i]-F[i-1] )*i )
就是那个11~12的解。
审题不清,罪过罪过~ |