v***o
发帖数: 51 | 1 谢谢楼上几位的真知灼见。MITBBS不正常，刚看到。
楼上讲的2/3也是我觉得头痛的问题，要有个很好的前瞻性。矛盾就是更多时候是没前
瞻性，因为有了问题也基本解决了。如何留有冗余又不失效率很没头绪。
另外个手头上的实际问题：如何解决用truncated distribution近似正常distribution
时两边tail过大？比如一个value function是个预期值，以mean为state variable，当
这变量取靠近两头的grid时，value function回overvalued or undervalued。还请高
人点拨啦。 |
|
I****M
发帖数: 143 | 2 另外个手头上的实际问题：如何解决用truncated distribution近似正常distribution
时两边tail过大？比如一个value function是个预期值，以mean为state variable，当
这变量取靠近两头的grid时，value function回overvalued or undervalued。
I faced similar problems. If the underlying stochastic process is stationary
, I usually use Tauchen Method for approximation; If it is a unit root
process, linear extrapolating the two ends of the state variables should
help. |
|
l********h
发帖数: 16 | 3 OK. some learning points, hope it helps.
"The impulse response decays much faster at the zero-crossings (
approximately as 1/t3 for t>>Ts) when compared to the "brick-wall" filter
(alpha=0). The rapid time rolloff allows it to be TRUNCATED in time with
little deviation in performance by theory" (credits to Pappaport)
"Increasing rolloff factor decreases the sensitivity to timing jitter, but
increases the occupied bandwidth."(credits to Pappaport)
"The time domain filter is NONCAUSAL. It must be |
|
h*****u
发帖数: 24 | 4 用VERILOG实现算法的时候，碰到ROUND函数，查了一下，有的是把real data type
truncated to integer data type,但是不能synethesis, 不知道有没有什么其他的解
决办法?多谢 |
|
F*M
发帖数: 1861 | 5 This is correct. In order to utilize the FFT size that are powers of two
for implementation simplicity and to maintain sub-carrier spacing wide
enough to prevent inter-carrier interference at high Doppler channels (LTE
uses 15 KHz sub-carrier spacing), oversampling has to be done, which results
in a larger initial bandwidth, which needs to be filtered (i.e. effectived
truncate the edge carriers) to fit the BW requirement. |
|
p******n
发帖数: 156 | 6 that's right. But I think CP is not guard period. Guard Period is to leave
the cyclic prefix part bland. Moreover,
CP can also prevent ICI (The ICI caused by truncation in time period. The
orthogonal property in frequency
requires infinity in time domain)
I forget two questions. See the following questions my answer is right or
not.
Q: CP can prevent the interference of previous signal. Does CP itself bring
interference to the current signal?
A: No, from derivation, at receiver, the current sign |
|
f*********t
发帖数: 37 | 7
Yeah, I agree with you at this point. From your original post, it sounds
like that you explain CP as a guard period.
(The ICI caused by truncation in time period. The
Why do you think orthogonal property requires infinity in time domain?
bring
is
I think your answer is correct to me, but maybe the interviewer want to
listen something about circular convolution :) or you use circular
convolution to explain will be clearer to them. |
|
l*******s
发帖数: 7316 | 8 1)
Dormand JR, Prince PJ. A reconsideration of some embedded Runge-Kutta
formulae. [Journal Paper] Journal of Computational & Applied Mathematics,
vol.15, no.2, June 1986, pp.203-11. Netherlands
Abstract
The RK5(4) and RK6(5) embedded Runge-Kutta formulae are reconsidered with
regard to enlarging regions of absolute stability while retaining satisfactory
truncation error norms. Results from standard tests for the above pairs are
presented in comparison with an efficient RK8(7) embedded formula. |
|
R*****s
发帖数: 28 | 9 The problem is like this:
Let X be a normal distribution with mean mu and standard deviation sigma.
Suppose M samples are drawn from X. However, only N ( N < M ) samples are
observable to us, with the property that these N samples have the largest
values of the set of M samples.
What are maximum likelihood estimates of mu and sigma?
Are they biased or unbiased?
Anyone has ideas? Can we use a truncated normal? |
|
c****n
发帖数: 2031 | 10 You mean you solve the semi-discrete equation analytically and then
truncate the summation to obtain a numerical solution? I think that's
only one of the spectral methods.
Anyway, I think we are far away from the original problem right now, hehe. |
|
e****c
发帖数: 183 | 11 比如一个 N （0，alpha)。另一个N (x, beta), 但是这第二个distribution 只在s 和
t 之间。这两个distribution 加起来，是什么样子？ |
|
b*****g
发帖数: 15 | 12 Let (X,O,mu) be a measure space. Let f be a non-negative measurable function
on X and N a positive real number. Does Chebychev's inequality bind at the
limit?
limit_N->infinite, Int{f dmu/X in mu(N)}/(N*mu(N))=1,
where mu(N)=mu{x in X/f(x)>=N}.
In fact, it is just a question about the conditional expectation with
truncation.
(If f is integrable, does it help?)
Thank you! |
|
S*********k
发帖数: 507 | 13 总是听到不区分 Taylor series 和 Taylor expansion . Series 是无穷项的求和。
expansion 一般是 truncated 的，对吧？ |
|
m******t
发帖数: 273 | 14 【 以下文字转载自 Statistics 讨论区 】
发信人: myregmit (myregmit), 信区: Statistics
标 题: find distribution paramters only by data mean and std. dev.
发信站: BBS 未名空间站 (Mon Mar 17 17:14:15 2014, 美东)
I need to estimate a truncated gamma distribution paramters (shape , scale).
But, I only know the data mean and std. dev. I do not know the dat set.
Given the mean and std. dev. of a data set from a trucnated gamma
distribution, how to find shape and scale of the distribution parameters ?
I know MLE may be sueful for ... 阅读全帖 |
|
m******t
发帖数: 273 | 15 I know that the shape and scale can be calculated from meand and std. dev.
I have tried this, but the QQ plot result is very bad.
I know this calculation works if it is gamma.
But, It is truncated gamma.
Thanks ! |
|
|
x***i
发帖数: 61 | 17 请问：什么软件可以画出各种形状的nanoparticle示意图？看到很多paper里面都有，
像 octahedral, cubic, truncated-octahedral... 像是很多原子堆砌成的形状，或者
就是简单的形状示意也可以的。谢谢！。
一般的化学软件只能画晶体结构啊！ |
|
S*****n
发帖数: 6055 | 18 首先我得承认我不懂TEM，然后恐怕你也没懂我的意思
我比较熟悉的统计particle size的方法就是chemisorption和EXAFS
这两种方法都需要作一个geometry或者说shape的假设
当然我们可以假设它是spherical的，但是对于lz说得3nm这么小的particle，因为已经
接近原子的大小，spherical的假设已经不精确了（因外表面不再能看成光滑表面），
通常是根据金属的表面性质做一些假设，比如“111面truncated的正20面体”，等等
我想请教的是，如果是用TEM做particle size的statistical analysis，是否也要做同
样的假设？ |
|
J****a
发帖数: 553 | 19 By my understanding, the DNA or amino acid sequence of the biosimilar
product need not be identical to the innovator product for a finding of
biosimilarity under the statute.There may be advantages to using a different
gene sequence that encodes the same protein.
"In general, FDA expects that the expression construct for a proposed
product will encode the same primary amino acid sequence as the reference
product. However, minor modifications such as N- or C-terminal truncations
that will not aff... 阅读全帖 |
|
N******o
发帖数: 25 | 20 I guess it's OK to have different gene sequence, but final protein sequence
has to be identical, as FDA mentioned. Minor mofification generally referers
to post-translation modification, including glycosylation, C-terminal Lys
truncation for mAb. FDA does gives small rooms for those known modifications
, especially glycosylation. Those have to be justifitable based on bioassay,
preclinical, even clinical data.
Overall, I believe no sequence variation acceptable in term of biosimilar.
different
... 阅读全帖 |
|
r******l
发帖数: 121 | 21 Now I see what the problem is.
The definition of NA = lambda_0/(n*D) is the NA of a Gaussian beam. In
common situations of focusing, we do not treat the beam as Gaussian beam but
a plan wave truncated by the aperture of the lens. In this way, the full
aperture of the lens is used and then one can achieve minimum focus spot
size. The spot size in the plane-wave incidence case is determined by the
Fourier transform relation (ref: Introduction to Fourier Optics, J.W.Goodman
).
If your situations is |
|
B*********h
发帖数: 800 | 22 ☆─────────────────────────────────────☆
killcy (崔色狼) 于 (Thu Jan 25 10:10:23 2007) 提到:
how to calculate something like 1/2 + 1/3 + 1/4 + 1/5 + ... + 1/n ?
thx
☆─────────────────────────────────────☆
wusuowei (wuxuowei) 于 (Thu Jan 25 10:33:00 2007) 提到:
google "hardmonic series"
☆─────────────────────────────────────☆
ThisTiger (aTiger) 于 (Thu Jan 25 10:41:06 2007) 提到:
If it is a series, it won't congerge; if truncated to n, is there a good
approximation for it?
☆─────────────────── |
|
D*****a
发帖数: 2847 | 23 有峰是因为truncation
mixed distribution 呵呵 |
|
l******i
发帖数: 1404 | 24 判断normal光看quantile distribution不行，得用formal test来证明，例如Jarque-
Bera Test。
我觉得楼上的truncation说法的很对，展开到infinity说不定就normal了。 |
|
M*****y
发帖数: 666 | 25 有没有高手来说说这道题怎么讨论？
两个人，A,B. A的年龄比Ｂ大。
For what length of time in their lives will A's age be twice B's age? (此处
的年龄说的是 truncated integer age). 还有一个assumption是两个人都可以无限地
活下去。
假设不限定具体的年龄数，该如何讨论这道题目呢？多谢！
我自己觉得无论Ａ，Ｂ多少岁，这个问题的答案始终是１年
不知道对不对，有什么合理的讨论方法 |
|
l******f
发帖数: 568 | 26 I used truncated normal for my simulation before due to asymmetric box
constraint like your [-8，10] |
|
w**********y
发帖数: 1691 | 27 I bet Compound Poisson should be general enough for your application.
1. Generate X_t, 0
2. Generate the Poisson Point process t_1, t_2,…t_k, which is the times
for jumps
3. Generate your jump size from a given distribution..
4. Modify your X_t..for t>=t1, Y_t = X_t + J1…until the last t_K
Glasserman’s book should have a clear step by step description.
When you move to the Levy process, it has two or three different simulation methods: one is truncated... 阅读全帖 |
|
v********r
发帖数: 15 | 28 统计的Ph.D. 论文做金融数学应用，面试了一些 Quant analytics
遇到几个问题想请教大家
1. 怎么样做保证beat S&P500 (if S&P up 1%, need to up more than 1%, if S&P
down 1% need to down less than 1%)，面试官给的hint 是，不用比S&P好很多，但是
一定要好
我之前读paper的时候，知道有outperformance analysis, 是given initial capital
， maximize outperformance probability. 如果是一定要beat, 用Super-
replication
面试官之后没有说对错，但是听语气好像不是他想要的，想问问这么个回答方向对不对
呢？
有没有人知道这个题该怎么答呢？
2. N stocks, N variables how to predict stock price in the future.
我答的是可以用 principle component analysis, (1) centralized,... 阅读全帖 |
|
c*******y
发帖数: 1630 | 29 我这里显示瓶颈在 上面一行
%time count, bins, ignored = plt.hist(s, 50, normed=True)
CPU times: user 36.4 s, sys: 751 ms, total: 37.2 s
Wall time: 37.2 s
3. |
|
c*******y
发帖数: 1630 | 30 贴到stackoverflow去问问，很多是写library的在上面晃
%prun -l 5 count, bins, ignored = plt.hist(s, 50, normed=True)
45928256 function calls (45910189 primitive calls) in 52.608
seconds
Ordered by: internal time
List reduced from 243 to 5 due to restriction <5>
ncalls tottime percall cumtime percall filename:lineno(function)
366366 14.342 0.000 14.342 0.000 weakref.py:267(__init__)
50050 6.618 0.000 15.825 0.000 artist.py:883(get_aliases)
50050 3.088 0.0... 阅读全帖 |
|
m******t
发帖数: 273 | 31 Thanks for your reply !
Would you please tell me how to get the detailed timing information and run-
time profiles for python ?
Thanks ! |
|
m******t
发帖数: 273 | 32 【 以下文字转载自 Statistics 讨论区 】
发信人: myregmit (myregmit), 信区: Statistics
标 题: find distribution paramters only by data mean and std. dev.
发信站: BBS 未名空间站 (Mon Mar 17 17:14:15 2014, 美东)
I need to estimate a truncated gamma distribution paramters (shape , scale).
But, I only know the data mean and std. dev. I do not know the dat set.
Given the mean and std. dev. of a data set from a trucnated gamma
distribution, how to find shape and scale of the distribution parameters ?
I know MLE may be sueful for ... 阅读全帖 |
|
p****u
发帖数: 2596 | 33 牛文转给你学习一下
发信人: yuxuxin (开水泡妞), 信区: Quant
首先，小伙子描述整个过程想证明自己很无辜，但其实仔细看他的描述还是能看出来他
说话的问题的。比如说那句“thank you for reaching out to me", 表面上看似乎没
什么问题，因为招人的经理找人跟他打电话的行动是可以说是reaching out to him.
但问题是他递简历在先，也就是他reaching out to the hiring manager在先。所以真
正礼貌的表达应该是先解释一下自己递简历的时候是认真的而且当时并不知道location
, 然后再 thank you for reaching out to me. truncate了前一个reach out 而只提
后一个 reach out 有点像强行把热脸凑冷屁股的感觉按到别人头上。在工作中，这种
事情一点都不trivial，如果你觉得trivial，那么我只能不客气地说那是你的职位还不
够，你的工作还是局限于干活。我曾经汇报给一个比我高3级的老板，亲眼目睹他是怎
么小心翼翼的跟不同部门说话，一句话要在脑子... 阅读全帖 |
|
s********n
发帖数: 12 | 34 on limited dependent variables. It covers truncated, censored and dummy
endogenous regression models. It even covers the Tobit + 2SLS model that I was
trying to solve couple of days ago.
Detail information of the book:
Name: Limited dependent and qualitative variables in econometrics
Author: G.S. Maddala
Press: Cambridge University Press
enjoy! |
|
|
G********t
发帖数: 334 | 36 the length of a char variable is defined when it first appears.
In this case, the length of job is 2, because of the statement job='fa';
Therefore, 'fa2' is truncated to 'fa'. |
|
c******j
发帖数: 270 | 37 The default length of numerical and character variables are both 8.
When input variables from a file,without specific formatting, char data will
be truncated to length 8. But numerical data will be taken in as long as it
is?
example
data test;
input animal $ mlgrams ;
cards;
hummingbird 90800000.87
;
run;
hummingbird is read in as hummingb but 90800000.87 is read in as 90800000.87? |
|
o******6
发帖数: 538 | 38 PROC IMPORT
datafile='K:\DATA ENTRY\P\Interviews\check.csv'
out=Pcheck
dbms=csv
replace;
getnames=yes;
datarow=2;
run;
那个dataset比较大，有4，5百个VARIABLE，而且很多VARIABLE是用来要SCORE的，
DATASET本身就需要CLEAN，已经花了我一些时间写MACRO来CLEAN和SCORE，比如有很多
DATE的VARIABLE，而DATE存的时候有些是8/ 9/2007，可是现在有个问题，我在SCORE的
时候发现有些VARIABLE被IMPORT后变成CHARACTER了，而且被TRUNCATE了，应该2位数的
变1位数了，上次EXCEL发现问题后CHANGE了WINDOWS REGISTRY WORK了，问题是我现在
还不知道怎么解决IMPORT CSV的问题，我不想用INFILE写，太花时间了，而且CODE又得
有变动，有没有人知道怎么解决这个问题呀，主要那个DATASET有好多DATA是MISSING，
应该也是 |
|
m******a
发帖数: 35 | 39 Regression Analysis when the Dependent Variable is Truncated Normal
Amemiya, Takeshi 1973
Estimation of relationships for limited dependent variables
J Tobin |
|
s*********e
发帖数: 1051 | 40 do you mean "negative binomial" ^_^?
standard Poisson definitely won't work, since your outcomes range from 5 to
12. however, truncated poisson might. |
|
T***9
发帖数: 73 | 41 大家有没有下面这本书的下载？
Survival Analysis: Techniques for Censored and Truncated Data
John P. Klein, Melvin L. Moeschberger
Hardcover, 2003
谢谢先！ |
|
c****s
发帖数: 63 | 42 Code:
data house.classmention;
set house.classmention classmention1;
run;
Log中出现的错误：
Warning: At least one data value for column M may have been truncated.
Maximum length of data encountered during input was 41.
请教怎么fix 这个问题，万分感谢！ |
|
|
x***i
发帖数: 135 | 44 Q54
Consider the following data step:
data WORK.TEST;
set SASHELP.CLASS(obs=5);
retain City 'Beverly Hills';
State='California';
run;
The computed variables City and State have their values assigned using two
different methods, a RETAIN statement and an Assignment statement. Which
statement regarding this program is true?
A. The RETAIN statement is fine, but the value of City will be truncated to
8 bytes as the LENGTH statement has been omitted.
B. Both the RETAIN and assignment stat |
|
|
i********w
发帖数: 2223 | 46 谢谢.
truncate 了以后,shape有点奇怪.我不知道该如何解释假设的合理性 |
|
j*****e
发帖数: 182 | 47 SAS help manu itself also offers a lot of examples on survival analysis.
Google "stat textbook" and "UCLA". The resulting website offers lots of SAS
codes for some of the most used stat books.
There are so many situations when collecting survival data. Left, right,
interval censoring. Left, right truncation. Each case corresponds to a
component in likelihood construction. It may be better to refer to a real
textbook, rather relying solely on a SAS book.
I never read Allison's book. If my memory |
|
h*****5
发帖数: 14 | 48 谢谢！
貌似GLIMMIX只能做MIXED EFFECT MODEL USING RANDOM STATEMENT，没有REPEATED
STATEMENT.
想找一个GEE macro，然后改写CODE， 插入一个TRUNCATED POISSON 分布。能否用SAS
实现？ |
|
t*********e
发帖数: 313 | 49 I have a variable called building in two files. The longest value of
building or the name of a building in one file is 36 bytes but has no format
or informat. In the second file, building has format/informat $17. When
merging the two files, the length of values for building is truncated to 17
bytes. So I guess it's because of the format/informat $17..
Can any friend help me how to avoid such trucation?
Thanks much!-James |
|
C******t
发帖数: 72 | 50 I don't know how to post the pdf file. By recogonizing the truncated normal
distribution, the solution is not very difficult. |
|
