s******s
发帖数: 13035 | 1 【 以下文字转载自 Statistics 讨论区 】
发信人: shakuras (doskey), 信区: Statistics
标 题: 问一个model优化的问题
发信站: BBS 未名空间站 (Tue Jul 3 17:12:34 2007), 转信
一般做一个full model, estimate了以后有几个参数可能不是significant. 然后怎么
建立优化的submodel? 显然不能只看p-value就把这几个参数去掉; 也不能全去掉, 然后
用submodel比, 因为去掉了significant的参数混着几个不significant的, 也能拿到比
full model更好的submodel. 难道只有建立一系列submodel, 每个比前面少一个参数,
一个一个比? 这样的话结果显然和去掉参数的顺序也有关系? 问一下大家一般是怎么做
的? |
|
f*******n
发帖数: 2665 | 2 “if there is a alignment model on top of these 50 submodels”你的意思是把50
个submodel做alignment以便使每个model score在数值相同时对应的风险也相同？ |
|
|
d***u
发帖数: 313 | 4 今天AMR，UAUA, LCC..等股票又在看涨。这些公司的fleet里，相当多数是波音飞机，
就做个波音民用飞机机型科普。
Boeing民用飞机分为短途的和中长途的（transcontinantal）
短途：
707 - out of production (oop) many years ago
717 - was assigned to MD airplane, oop many years ago
727 - oop many years ago
737
757 - oop few years ago
中长途：
747
767
777
787
飞机size的从小到大排序为 - 707，717，727, 737，757, 767, 787, 777/747
除了这些主要models以外，每个主要model下还有submodels. 其中以737最多。
737-300，400，500 （统称 classic, oop few years ago)
737-600,700,800,900 (统称New Gen or NG)，737-700，800是这些年来最畅销的机型。
在中长途飞机 |
|
p***p
发帖数: 559 | 5 It takes me 2 weeks time, class after class.
The Project manager must be a fan fo aircraft
Catalina is nickname of PBY 5, a famous amphibios plane
Tomcat is the nick of F14
Coyota is the nick of a submodel of P3 Orion
Apache is nick of AH64 Helicoper |
|
l**********1
发帖数: 5204 | 6 需要genome 生统的背景
比如 H 大 以下 两lab (NB: one now is at Stanford)
//biosun1.harvard.edu/complab/
的至少one first author a paper 那样的背景
sunny 再去修个MD 如何
one mentor:
//www.hsph.harvard.edu/faculty/frank-hu/
cited from last week Cell cover article:
pp393
>With respect to the first question, replication initiation occurs
>as DnaA protein monomers bind or unbind stochastically and
cooperatively to form a multimeric complex at the replication
origin (Figure 4B, top) (Browning et al., 2004). When the complex
is c... 阅读全帖 |
|
l**********1
发帖数: 5204 | 7 only used ODE (Ordinary differential equation) as Simulation Algorithm
Simulation Algorithm
The whole-cell model is simulated using an algorithm comparable to those
used to numerically integrate ODEs. First, the cell variables are
initialized.
Second, the temporal evolution of the cell state is calculated on a 1 s
timescale
by repeatedly allocating the cell variables among the processes, executing
each of the cellular process submodels, and updating the values of the cell
variables. Finally, t... 阅读全帖 |
|
|
r********r
发帖数: 281 | 9 转贴的，请直接联系公司。
I am currently looking to fill 2 senior engineering roles at my company,
Echogen, which is an emerging clean tech start-up company, based in Akron
Ohio.
The first role is a Senior Systems Engineer, which would provide the
opportunity to be a technical team lead in a new product introduction
environment.
In additon, I am looking for a Senior Engineer to lead our Thermodynamic
Cycle Design Team. A little about the role is below.
I was hoping you might be interested in one of these rol... 阅读全帖 |
|
b*****a
发帖数: 905 | 10 lz自己挑著看吧。
## Make a ROC plot to see the ration of the rate of false positives to the
rate of false negatives
library(ROCR)
ip.glm<-glm(votesum ~ clint96 + partyid + aflcio97 + ccoal98, data=ip,
family=binomial(link=logit))
summary(ip.glm)
names(ip.glm)
pred.object<-prediction(ip.glm$fitted.values, ip.glm$y)
perf.object<-performance(pred.object, "tpr", "fpr")
# use add=TRUE as an argument to the plot() function if you want to overlay
additional ROC plots.
# Compare the model with submodel (Democr |
|
c********d
发帖数: 253 | 11 来自主题: Statistics版 - 一个理论题 likelihood ratio requires one model to be a submodel of the other one. In
your case, the two are seperate models. Baozi, plz, hehe:) |
|
s*********e
发帖数: 1051 | 12 if there is a alignment model on top of these 50 sobmodels, then it is
possible that the probability from different submodels might represent the
same risk.
logistic |
|
