l*****y
发帖数: 4887 | 1 乱说一个，你按照maximum likelyhood搞的unfair coin with prob=p
难道不是现在
p^3(1-p)^0
基本p的maximum likelyhood estimate就是1了
也就是你已经没希望了？
拯救一下我已经忘掉了的数学吧 |
|
a***r
发帖数: 420 | 2 "Then, we update the haplotypes for each individual in turn by using the
current set of haplotype estimates for all individuals as templates and
sampling S proportional to the likelyhood L(S|G)∝P(G,S).A new set of
haploytpes for an individual is then defined according to sampled mosaic."
其中， S是HMM中的hidden states， G是observations
P(G,S)作为S和G的joint probability，可以由markov chain基本公式得出，认为已知
关键是“sampling S proportional t the likelyhood L(S|G)∝P(G,S)”，不明白是
咋个sampling的过程，大致觉得是说对每一个individual来说（每个G），相对 ”th |
|
f*******y
发帖数: 258 | 3 This is the 2nd family been cruelly killed in United States.Owning gun may
be a solution, as by owning guns, the family would be more vigilant than the
families without the protection of guns. As for the accidental shooting, I
don't think the likelyhood is higher for these gun-owning families.
Control what you can control and prepare for the worst. I own guns and I
encourage my wife and my kids to know how to use guns safely. In the face of
crime, if you can not prevent the crime from happening,... 阅读全帖 |
|
f*******y
发帖数: 258 | 4 This is the 2nd family been cruelly killed in United States.Owning gun may
be a solution, as by owning guns, the family would be more vigilant than the
families without the protection of guns. As for the accidental shooting, I
don't think the likelyhood is higher for these gun-owning families.
Control what you can control and prepare for the worst. I own guns and I
encourage my wife and my kids to know how to use guns safely. In the face of
crime, if you can not prevent the crime from happening,... 阅读全帖 |
|
f*******y
发帖数: 258 | 5 This is the 2nd family been cruelly killed in United States.Owning gun may
be a solution, as by owning guns, the family would be more vigilant than the
families without the protection of guns. As for the accidental shooting, I
don't think the likelyhood is higher for these gun-owning families.
Control what you can control and prepare for the worst. I own guns and I
encourage my wife and my kids to know how to use guns safely. In the face of
crime, if you can not prevent the crime from happening,... 阅读全帖 |
|
c*********d
发帖数: 9770 | 6 原来probability和likelyhood在数学里都是有严格定义的啊，不能混用。 |
|
k***n
发帖数: 3158 | 7 I have said it clearly
if Romney does what he says, he would be marginally better than Obama
Otherwise, he is just another Obama and this likelyhood is higher IMO |
|
a*****h
发帖数: 484 | 8 Thanks for the very good points! I guess one should consider the people to
work with and the likelyhood of getting tenured there...
(
good |
|
|
l*******h
发帖数: 151 | 9 今天先说BF的经历吧。交代下BF的背景，中部数学系精算小硕，过了４门考试，无实习
经验。４月底毕业，从去年１１月开始找工作。由于之前他简历上的毕业日期写成了２
０１０年，从１２月到３月间没有一个电话面试。４月一个大公司电面，第三轮时因为
解答一个DATABASE题目的速度太慢，止步于此。
７月通过内部员工推荐，开始面试一家大型MORTGAGE公司。总共已经面了４论，２轮电
话２论ONSITE。面试的职位分别是DATA ANALYST和DATA MODELER。
细节如下：
第一轮：HR电话面试。很基本的问题，２０分钟－３０分钟，教育背景，期望工资水平
等。之后安排了第二周TECHNICAL电话面试。
第二轮：TECHNICAL电话面试。３０分钟，主要问EXCEL和SQL。其中一个问题是，SQL里
有２种JOIN，解释其区别和作用。另外还有问BF平时有用EXCEL做过什么PROJECT。
第三轮，ONSITE－１。DATA ANALYSIS小组，８个人面他。基本没有TECHNICAL问题，８
个人问的问题都很随机。比如，你怎样估计CHICAGO总共有多少PIANO TUNER？你最拿手
... 阅读全帖 |
|
t*******n
发帖数: 4445 | 10
坏。
以现在进行时长着的，估计还要继续长。最近没有长的，仔细问医生：will grow还
是may grow, and does he have an idea about the likelyhood.
中文没时态就是费劲 :-(
可以严肃地问一下医生可能的和最坏的结果。天下没有100%安全的手术，医生也不能
重描浓写，动不动拿0.001%的概率吓唬人啊。你要实在担心可以不做，等出毛病了再挖
掉，不过可能困难点外加多麻一次。 |
|
g***r
发帖数: 285 | 11
rheumatoid factor is positive in many patients such as mixed cryo, lupus,
scleroderma, and even healthy patients, so by itself it doesn't help in
diagnosing rheumatoid arthritis,
you need to ask her morning stiffness how long does it last, morning
stiffness occurs in almost all inflmmatory arthropathy but if it lasts more
than one hour, then the likelyhood of rheumatoid arthritis is quite high,
you do not need to be hospitalized at all, there are different medications
you can try, |
|
g***r
发帖数: 285 | 12
painless jaundice and pancreatic mass in an elderly is almost always
malignant, you can check tumor marker, ca 19-9, if it's very very high, the
likelyhood of pancreatic cancer will be very very high,
in that case it's better not to get a biopsy to establish the diagnosis, at
age of 90, agree with gordonj that better just leave him be, keep him
comfortable, since you will not treat him even if he was diagnosed with
cancer, |
|
g*****j
发帖数: 1211 | 13 Headache is a fairly common symptom that can be caused by a huge list of
problems ranging from benign causes like tension headache to grave causes
such as meningitis. When it comes to diagnosis, as you could imagine,
doctor will have to consider the most common causes first and at the
meantime watchful for those acute and serious causes.
Based on your history, I will consider sinus infection first as well. You
should understand antibiotics does not guarantee cure. Doctor probably give
you a r... 阅读全帖 |
|
w**********9
发帖数: 242 | 14 anything would be possible. no matter what, a thorough investigation will
know what it is. I think the likelyhood is small because of infections
associated. |
|
B******1
发帖数: 9094 | 15 A well written article.
However,
neither am I critizing any group -> criticizing
predictable and quite child -> quiet
kindergarden -> kindergarten (the cardinal sin on this board)
extracurriculum activities -> extracurricular activities
martial arts, ect. -> martial arts, etc.
But the likelyhood is very low nowadays. -> likelihood
Only 1 in 1000 children are disabled or die from accident. -> 1) would
be disabled or die 2) is disabled or dies
childhood desease -> diseases
It is a highly personal |
|
v*****i
发帖数: 27 | 16 then to be millionare, this company has to be worth 100M, do your own math,
consider the likelyhood in the next 1/3/5 years and how long you can commit
to. do they already own a market share or still a prototype, that makes a bi
g difference too.
100K |
|
k****y
发帖数: 4083 | 17 you could try, but please note the likelyhood of sucess is <1% |
|
w****b
发帖数: 623 | 18 Finessing H succeeds when H is 3-2, and HK onside, a 31% chance. The chance of
finessing D is more complex. You succeed when DK onside, D3-3, or 4-2, but the
person holding short D does not have Hxxx. It's actually slightly better that,
since if you play DQ, if East has DKx Hxxx, it's hard for him not to cover.
Then even he has Kxxx, there's a likelyhood that he'll cover (you might have
AJxx). So you will make this most of the time even when East had AQxxx Kx Kxxx
xx.
So I'd say finesse D. |
|
f*****g
发帖数: 15860 | 19 i don't have an odds calculator here i was laughing to see the cards and
flop... AK purely was drawing to dead, and i was magically still alive!
have seen enough KK beat AA, AK beat KK and AK beat AA (high to low
likelyhood)... this one is still an eye opener to me, haha, collected in my
weird hands collection. |
|
f**********d
发帖数: 47 | 20
the moment that you discover yourself 非常非常喜欢他 has put you into
difficult position. notice i didn't use the word dangerous which means lose
yourself and your ground.
you care more about the relationship and will indeed enjoy and suffer more.
i feel like you post here because you started to wondering and questioning
the likelyhood of the continuation of this relationship,
the truth is that no one will have a real solution, every relationship has
its uniqueness.
there are certain things we can con |
|
j******u
发帖数: 7 | 21 fiml the whole system like to estimate each equation by OLS?
如果没甚麽区别,是不是FIML the whole system 比较好?
一见matix 就头疼的说 |
|
m****i
发帖数: 1 | 22 I guess FIML estimates are the same as running reduced form. It
yields consistent estimates. It's definitely different from OLS.
FIML is better. |
|
g*****r
发帖数: 4 | 23 Just finished my econometric comp. FIML is asymptotically
equivalent to 3SLS without constraint on the variance matrix.
We know that if there exists heteroskedasticity or in the case
of simultaneous equation model, 3SLS is preferred, and thus FIML. |
|
j******u
发帖数: 7 | 24 is there some software could estimate the parameter by fiml?
you mean it is equal to 3sls, so is there some software do 3sls?
i knew eview can, but i think it is too complicated.
其实,我不明白FIML的本质的说.
I knew a little about 2sls, is 3sls like 2sls? |
|
g*****r
发帖数: 4 | 25 If your sample size is big, FIML is equivalent to 3SLS, and
LIML is equivalent to 2SLS. The difference between 3SLS and
2SLS is the M matrix. I wonder if you know Q=1/2*H'MH. In
2SLS, M=inv(1/T*W'W), where W is the instrumental matrix. In
3SLS, M=inv(variance matrix/T*W'W). These are all in
homoskadastic case. If the variance matrix is not diagonal
with 1's, 2SLS is not efficient, ie you have to use 3SLS. If
you are sure there exists heteroskadasticity, you have to
construct the M matrix |
|
m**********3
发帖数: 706 | 26 Again, I think in the real exam you have to pick the most possible diagnosis
. For example, a pt. with headache and his csf is full of WBC, is it p.
strep or N. meningitis. ? I think if the pt. has no rash, p. Strep is a
better choice,although both are very likely, and N. Meningitis will not
necessarily show rashes.
Hope you guys can understand what I mean in the real test. The above Q if
the pt. has no rash, maybe VIPoma, because it is MORE LIKELY that
glucagonoma will have rash. (some may not ... 阅读全帖 |
|
p********0
发帖数: 186 | 27 Hi,
For GARCH model
\sigma_{t} = a + b * \sigma_{t-1} + c* \return_{t}, Can I run a linear
regression
to get the parameter? Can I use the regular linear regression since the \
sigma_[t-1} is
a response variable also a X variable? CAN I USE THE REGULAR regression?
Or to use a Maximum likelyhood function/get a derivative first, plug in the
\sigma_{t-1}
to \sigma_[t}. In this way I can get a parameter estimate? |
|
a****n
发帖数: 230 | 28 一直以来潜水在这个版，从一无所知到选择修课，准备面试，学到了很多东西。offer
老板说今天寄出来，昨天开始就难抑心中欣喜。虽然不是完全quant role，但是在那个
group里面，算是一个对没有经验的人的很好起始点，我已经很满意了。
发一些面经吧：
电面：
第一个人：全是简历上的，每一条刨根问底（搜狐栏目组）问，幸亏之前吸取了一次失
败的电面，所以准备还算充分。
code test和几个小问题：非常简单。
第二个人：一半时间简历。然后说这种面试按例要来一些puzzle吧，就出了经典的100
层楼扔两个鸡蛋，然后我当无知状分析了一下。接着他说：那假如三个鸡蛋呢？没做过
，顿时手心冒汗，心跳加速，一顿分析狂狂说，结果我还没怎么整明白呢，他说good
enough.汗。。。。。
onsite: 5 senior people (all are very nice)
1. 还是简历，project，research
2. linear regression： derive the equation, and relate it to maximum
likelyhood.
3. what |
|
t*****j
发帖数: 1105 | 29 就看这个(1/a)^n，要maximize这个就得minimize a
since a》=max(Xi)，所以maximum likelyhood estimation就是max(Xi) |
|
m******a
发帖数: 35 | 30 estimation is as following:
y*=x'b+e, e follows N(0,sigma^2) , normal distribution
y*
a1=
y*0
How could I get the Ln[G(yj|Xj,a,b)] in order to implement the ML estimation
for a1 and a2?
Could anyone please help me ?
Many thanks |
|
|
n*****1
发帖数: 172 | 32 this question is too broad... |
|
d******e
发帖数: 7844 | 33 如果对于regression来说，那么：
OLS不能用于Generlized linear Model。
对于linear model来说，二者没啥本质区别。 |
|
w****j
发帖数: 6262 | 34 能展开说一下，ols为啥不能用于generalized linear model么？
多谢。 |
|
o****o
发帖数: 8077 | 35 because of the 'O' in OLS |
|
H******r
发帖数: 2879 | 36 ols在没有missing data的时候是ML。 |
|
s*****r
发帖数: 790 | 37 are you sure?
Maximal likelihood estimate, there is a word 'likelihood' in it, which means
it requires distribution. Just under the normality assumption, it is OLS.
Otherwise, it may not be. |
|
a***r
发帖数: 420 | 38 我其实就是用这个格式存的：）在原帖里把问题改了一下，想显得general一点（呃，
似乎有点多余。。。）
不过，你说的把所有结果都导出到这个X矩阵然后再找，其实就是songkun的思路啊~
嗯，我关于那个1500*1500说得不够清楚
数据结构是这样的
obs Y x1 x2 x3... var1500
1 1 1 0 2 ...
2 0 2 1 0 ...
...
Y是0/1 categorical variable
x可以看成numeric variable
然后我想知道的，是对于所有x的pairwise的组合，增加一个interaction term对原
model的improvement，即
nested model： y=x1+x2
full model：y=x1+x2+x1x2
然后对每对做log likelyhood ratio test，得到相应的chisq和P
因为是preliminary study，想得到一个尽量inclusive的list |
|
l*********s
发帖数: 5409 | 39 Of course. and I don't see the practical distinction between likelyhood and
sample. |
|
m*******e
发帖数: 90 | 40 发现很多情况下，log likelyhood的最大值都不在这个区间里啊。
大家能指点一下吗？谢谢！ |
|
c***z
发帖数: 6348 | 41 @longtian大牛：
Issue 2 能展开说说吗？我看了一下optim的代码，它会call一个C程序叫做C_optim，
好像收敛条件和迭代数量得在那里改？有点头大，很久没有弄C了。
@zhanghan大牛：
原作者没有pass gr，不知道从那个超复杂的likelyhood弄个gr会有多麻烦？
多谢各位大牛！ |
|
