p******s
发帖数: 229 | 1 接上面的问题。
要求得每个医院 estimated probability and its 95% CI of event y, 在 proc
logistic regression 加一个什么样的output statement 可以得到呢？
谢谢大家帮助新手！！
by statement will do it:
proc sort data=all; by hospital; run;
proc logistic;
by hospital;
model y=x1 x2 x3;
output .........;
run; |
|
o******6
发帖数: 538 | 2 ☆─────────────────────────────────────☆
smalltomato (我爱番茄鸡蛋面) 于 (Tue Feb 10 11:47:11 2009) 提到:
没有找到相关的文献。
一个统计Ph.D说，不一定。
但在一些文献里只是不假证明地claim说，Bayes estimates are more accurate.
☆─────────────────────────────────────☆
auraxr (AURA XR) 于 (Tue Feb 10 13:19:56 2009) 提到:
More Accurate是怎么定义的？
☆─────────────────────────────────────☆
oloolo (高大全不沾男人) 于 (Tue Feb 10 13:23:25 2009) 提到:
more efficient?
☆─────────────────────────────────────☆
smalltomato (我爱番茄鸡蛋面) 于 (Tue Feb 10 13:30:24 2 |
|
D******n
发帖数: 2836 | 3 let d be an estimator ,S be sample
deterministic means Var_d(S) (d(S)) = 0;
variance means Var_S (d(S)) !=0 ;
it is very simple, from S to d(S) it is deterministic,not probabalistic,
variance of d(S) is in terms of S, S is random so d(S) is also random |
|
S******y
发帖数: 1123 | 4 I have a one-variable data set in R.
The plot of histogram of my numerical variable suggests an inverse
gaussian distribution.
How can I obtain best estimation for the two parameters of inverse
gaussian based on my data?
Thanks. |
|
x*******i
发帖数: 1791 | 5 i remember there is transformation form to estimate two parameters of
inverse gaussian. i think you can do gibbs.
something like two roots of the function.... |
|
c*****r
发帖数: 2608 | 6 a data fitted to a mixed model, time as the random effect, and x1 and x2 be
two fixed affect, with coefficients b1 and b2.
question is how to estimate the confidence interval:
1. F = arctangent(b1/b2)
2. A = b1/sin(F)
thanks |
|
P****D
发帖数: 11146 | 7 Delta method, perhaps? I just cannot see the point of estimating F and A. |
|
f**n
发帖数: 401 | 8 Assume a n-dimensional random vector x = [x_1,...x_n]' comes from a
multivariate normal distribution. The unknown mean vector and covariance
matrix of x are subjects of our parameter estimation and we would like to
get it from the following relationship:
There are k samples of x, denoted by x1, ..., xk. However, we cannot observe
x1,...xk. Instead we can only observe another n-dimensional vector yk = [y_
{k,1},...y_{k,n}]'. We define z = y \cdot x. i.e., z is the doc product of
vectors x and y. |
|
l*********s
发帖数: 5409 | 9 well, our teacher mentions something like consistent estimation does not
require existent variance... is that possible? What are the conditions to
ensure consistency? |
|
l*********s
发帖数: 5409 | 10 I often read claims that "xxx regression estimator has desirable features
such as shrinkage " blablabla, what is so nice about shrinkage anyway? @__@ |
|
A*******s
发帖数: 3942 | 11 I saw some conclusions many times like:
sample estimations of larger eigenvalues (of covariance matrix) are biased
high, and those of smaller are biased low.
Could any big bull tell me why? thanks in advance and baozi will be sent
later |
|
A*******s
发帖数: 3942 | 12 No. i saw it on Freidman's paper "Regularized Discriminant Analysis". Right
after the formula (18) you can see his description of this problem.
"This shrinkage has the effect of decreasing the larger eigenvalues and
increasing the smaller ones, thereby counteracting the biasing inherent in
sample based estimation of eigenvalues."
I recall I saw similar statements before but I don't know where it is from.
re |
|
s*****9
发帖数: 108 | 13 这方面的paper还挺多的。说说我的想法:
比方说做PCA, 如果sample size n和参数空间维数p相比不算太大的话。那么估计出来
的头几个PC所代表的variation会比其实际的要大一些吧（co-linearity等等因素），
因此最大的几个eigenvalue就会被over-estimated。总的variation的估计不会偏差多
少，那么最后几个PC所在的variation就会被压缩的很小。最小的那个eigenvalue经常
会被shrink到0，导致covariance matrix不可逆。要是n >> p ，应该不存在这个问题
吧。 |
|
I*****a
发帖数: 5425 | 14 Yes this is intuitively understandable. I think when the sample size is so
large that the data are not sparse in any dimension, then the estimation is
not problematic.
I just checked Friedman's paper, and he said the biases are most severe espe
cially when the eigenvalues are close to each other. I don't quite get that.
Do you know the reason for this ? |
|
F****n
发帖数: 3271 | 15 MLE needs a pdf. In your case, g(X) is pdf, but you are actually estimating
g(X(V)), which is a new function of V. Make sure it is also a pdf of V.
.v
where |
|
n**f
发帖数: 121 | 16 Thank you very much for the reply. I guess this is the problem.
However, the distribution of V is very hard to derive. Is there anyway to
get around that? Like method of moments?
Again many thanks!
estimating |
|
n**f
发帖数: 121 | 17 Thank you for your patients comments. Now I know the issue is with modeling
rather than estimation.
theta,
method. |
|
|
e*****u
发帖数: 67 | 19 I have built a logistic model to predict binary outcome.
My prediction for data points from validation set is p-hat (predicted
probability of event), i.e., a point estimate.
Is there a way to build a confidence interval for my logistic prediction,
just like in OLS linear model prediction?
Thanks. |
|
o******e
发帖数: 1001 | 20 如果假定一堆数据符合同一正态分布，那么它的MLE很简单，不需要计算maximum
likelihood, 我们就可以知道期望值是样本均值，方差是样本方差，而且如果把样本均
值和方差放到maximum likelihood的那个公式了，应该能够得到最大的likelihood。
问题是，如果这堆数据本身并不是特别符合正态分布,有没有这种可能，用R解出来的
maximum likelihood estimator和样本均值，样本方差很不一样？这时候把样本均值方
差替入maximum likelihood的式子得到的likehood并不是最大的？ |
|
S****Y
发帖数: 4634 | 21 问一下，怎么样用SAS或者STATA estimate logistic regression,
with clustered standard error in two dimensions.
谢谢！ |
|
q****8
发帖数: 1 | 22 Does anyone know how to use drc package to estimate IC50? Which model is
normally used? I searched from internet that IC50 is the dose that
corresponds to half of the upper and lower limit of the curve, does that be
the same as ED50 if I use LL.3 in drc package?
I am confused about the definition of IC50 calculation. If a drug is very
resistant, normally the IC50 should be very high, however, if we use the
definition that IC50 is the dose response to half of the upper and lower
limit, it should ... 阅读全帖 |
|
l*******3
发帖数: 186 | 23 做maximum likelihood estimation，用optimization怎么都的不出正确的数值。现在
是做simulation阶段，总共10个parameters，如果固定其他parameters，只让一个
parameter变，那么likelihood function确实总在true parameter valuealuede地方是
最大的，但是所有parameters放在一起optiomize likelihood function，就怎么都的
不到真实parameter value。到底是怎么回事呢？大侠帮帮忙啊，谢谢！ |
|
M*******r
发帖数: 165 | 24 就是最普通的probability distribution, 比如normal, 客户要求用CV检验这个
distribution的estimation是否valid，请高手指点下，这样做合理吗？我认为CV更多
是用来检验prediction model, distribution的test更多是用hypothesis test来着，
比如chi square test。 |
|
g**u
发帖数: 205 | 25 借地方问一下，如果用proc mixed做出来的LS mean显示是non est，那说明数据有什么
问题啊？能查到的资料上都写non est是指L matrix fail to test estimability，但
这说明了什么啊？实在想不通。 |
|
l**********o
发帖数: 2 | 26 ods trace on; find out the parameter estimates, then use ods output |
|
o******e
发帖数: 1001 | 27 多谢指点！能讲的稍微详细一点吗？我对Kernel Density Estimation 没有多少概念。 |
|
p*********g
发帖数: 453 | 28 最近想做一些Maximum Likelihood Estimation，没有标准的package可以用，所以就用
R写。但是有一些不清楚的地方，比如用那个function，initial value该怎么设置。
谢谢！ |
|
y*c
发帖数: 247 | 29 There is no interactions in the model. Like linear regression, you plug the
means for the baseline covariates in the model in order to estimate group
means for the main treatment effect. But I have a time-dependent covariate
now. what should I plug in the equation? The average of the time-dependent
variable (infection during follow-up) across all visits and all patients?
what would be more meaningful? Hope somebody can help me.
|
|
y**3
发帖数: 267 | 30 thanks for the reponse! Can you explain why? I guess the interaction term of
height and weight is insgnificant. Can we assume height and weight are
independnet, and calculate the parameter for BMI
Unfornately I dont have the original dataset. SO I can refit the original
model using BMI. And The goal is to validate the original model estimated
coefficients |
|
y**3
发帖数: 267 | 31 Thanks!
Yes,it makes sense!
I tried out another data set from a different product. The fitted BMI
coefficient is close to the one calculated from those of height and weight,
but all the other estimated coefficient also changed dramatically |
|
a******e
发帖数: 119 | 32 I estimated the mean and variance of two latent variables through two groups
of data. I can't use the data to do hypothesis testing, because I am
interested in the latent variable. Is there a way to test the whether the
two latent variables are significantly different? |
|
f***8
发帖数: 571 | 33 【 以下文字转载自 DataSciences 讨论区 】
发信人: f0008 (f0008), 信区: DataSciences
标 题: Google CasualImpact - R package for estimating casual effect in time series
发信站: BBS 未名空间站 (Thu Sep 11 19:24:46 2014, 美东)
Google刚刚发布了一个R package: CasualImpact. 估算时间序列里的因果关系。有没
有高手愿意点评一下？
http://google-opensource.blogspot.com/2014/09/causalimpact-new- |
|
|
D****G
发帖数: 284 | 35 如果客户要求在50,000中选出 top20 risky target, 那岂不是不能根据estimated
probability of default 来排序取最大的20个？这样的话，模型没法用啊，project的
目的就是找出top 20 risky的target.
minimal |
|
b********r
发帖数: 37 | 36 大牛，你怎么解释kernel density estimation? |
|
T*********s
发帖数: 17839 | 37 Estimated Flows to Long-Term Mutual Funds
Millions of dollars
12/14/2011 12/21/2011 12/28/2011 1/4/2012 1/11/2012
Total Equity -5,917 -4,690 -5,085 -9,365 1,434
Domestic -4,229 -2,807 -4,111 -7,079 753
Foreign -1,689 -1,883 -974 -2,285 681
Hybrid 752 172 460 454 1,951
Total Bond 5,545 2,272 2,073 3,298 7,868
Taxable 4,684 1,090 1,079 2,120 6,124
Municipal 861 1,... 阅读全帖 |
|
T*********s
发帖数: 17839 | 38 Estimated Flows to Long-Term Mutual Funds
Millions of dollars
12/14/2011 12/21/2011 12/28/2011 1/4/2012 1/11/2012
Total Equity -5,917 -4,690 -5,085 -9,365 1,434
Domestic -4,229 -2,807 -4,111 -7,079 753
Foreign -1,689 -1,883 -974 -2,285 681
Hybrid 752 172 460 454 1,951
Total Bond 5,545 2,272 2,073 3,298 7,868
Taxable 4,684 1,090 1,079 2,120 6,124
Municipal 861 1,... 阅读全帖 |
|
M*******a
发帖数: 870 | 39 哪里查比如mw的first quarter eps estimate是多少？ |
|
|
t*******o
发帖数: 1464 | 41 果然， TLB的sales miss了， EPS beat不是销售项带来的。虽然miss的不多...
Sales at the Hingham, Massachusetts-based retailer have fallen for three
straight years as consumers conserved cash to cope with the economic slump.
Revenue of $320.7 million for the three months ended May 1 missed the $324.6
million average estimate of five analysts compiled by Bloomberg. |
|
|
y*****l
发帖数: 5997 | 43 【 以下文字转载自 Stock 讨论区 】
发信人: djinwa (djinwa), 信区: Stock
标 题: Re: Estimate Near Term Target for your pick
发信站: BBS 未名空间站 (Sat Mar 5 19:54:30 2011, 美东)
MRVL 15.09
OPK 4.02
MWW 15.46
NYT 11.5
RAD 1.36 |
|
y*****l
发帖数: 5997 | 44 【 以下文字转载自 Stock 讨论区 】
发信人: djinwa (djinwa), 信区: Stock
标 题: Re: Estimate Near Term Target for your pick
发信站: BBS 未名空间站 (Sat Mar 5 19:54:30 2011, 美东)
MRVL 15.09
OPK 4.02
MWW 15.46
NYT 11.5
RAD 1.36 |
|
b**********k
发帖数: 1262 | 45 盘后跌了
estimates-q1-view-beats/ |
|
a***n
发帖数: 5665 | 46 that sounds reasonable to me. but since GEICO wont pay a cent, u can take it
to
the other place for an estimate. how much is ur insurance premium?
and
guaranteen |
|
l****y
发帖数: 3925 | 47 you mean to the dealer certifed place? I do not know which one is gonna do a
better job. Maybe dealer certified place has more experience with BMW cars.
My premium is about $440/6 months, 500 deductible
Geico said no matter where I go, the cost will not be more than what they
estimate, they will deal with body shops about the price. I may find a
cheaper one.
it |
|
l****y
发帖数: 3925 | 48 right, the price is fine. I can not save much on this. I just want a better
job.
The place I went to today for estimation seems ok, they said they will make
it
exactly the same as what it was. GEICO will ganranteen for the work. But
the certified BMW is not certified by GEICO, kao |
|
f***s
发帖数: 533 | 49 take it to the boday shop with the estimate, if they feel it will cost more,
they will contact GEico. Geico (or other insurance cos) always underestimate
the damage.
a
.
take |
|
d***e
发帖数: 710 | 50 Let's estimate the weight. How many pounds each person
need to carry. I hate carrying heavy stuff. |
|
