c******d 发帖数: 1150 | 1 【 以下文字转载自 Statistics 讨论区 】
发信人: caldavid (认识), 信区: Statistics
标 题: 有没有什么软件可以实现用多种model来fit data的?
发信站: BBS 未名空间站 (Thu Feb 12 22:38:37 2015, 美东)
matlab我知道的能一个model一个model试,但是需要试的model太多了,想问问有没有
什么软件能够只要输入自变量和因变量就能使用不同model来模拟,算误差的。。
谢谢 |
|
j****i 发帖数: 68152 | 2 【 以下文字转载自 Automobile 讨论区 】
发信人: qgod (Q神), 信区: Automobile
标 题: [BSSD]我组里一个刚毕业小中要买Model X, 经我开导后买了GLC
发信站: BBS 未名空间站 (Mon Nov 12 17:51:59 2018, 美东)
我组里一个今年刚校招进来的direct report给我的小中,本科TOP2、屁挨着地半路转
[email protected],国内家住北京东城区四合院。 他进来时拿的是32万的标准包裹(其中12
万sign on)。
上个月某一天我看他在特斯拉官网配置Model X的页面,我就问他是不是要买车。 他说
准备年底前买辆Model X或卡宴。
我于是问他 - 你还是single,住的也是公司免费提供的走路到office 5分钟的luxury
condo(房租按市场价的话至少5~6千一个月,地点NYC),要买车干啥子?????
答曰:周末和女朋友出去玩开开,另外父母春节期间来美,想开Model X去接。
经过我悉心劝导,他同意委屈一下买辆奔驰或宝马。
然后我就给推荐了GLC Coupe,... 阅读全帖 |
|
发帖数: 1 | 3 【 以下文字转载自 Automobile 讨论区 】
发信人: rcam (八戒悟空), 信区: Automobile
标 题: model 3 车主冬天的经历
发信站: BBS 未名空间站 (Sat Feb 9 00:11:01 2019, 美东)
https://www.zhihu.com/question/63220339/answer/582868280
这两个充电站隔了200mi的路程,中间一个超充都没有。当天早上我开着强行充到了
290mi的Model 3上路了,车上的导航告诉我开到page充电站还能剩百分之二十左右的电
。因为绝大部分公路被冰雪覆盖,经常连路上的线都看不见,所以一路平均速度只有50
迈,按理来说要能再多开几十mi的。然而等我午后离开了第一个旅游景点(不到1/3的
路程),发现预计到达Page充电站的剩余电量已不足10%,感觉按照这个预计剩余电量
的线性折损速度开到Page只怕是只会剩-10%的电量了。于是我开始查询路途中的第三方
充电站,毕竟冰天雪地里龟速行驶能少开一点就少开一点,中间充几个小时电也无所谓
了。
不料美国的几大电动车充电网络在这段路上一... 阅读全帖 |
|
l****z 发帖数: 29846 | 4 The observed response of the climate to more CO2 is not in good agreement
with predictions.
By WILLIAM HAPPER
During a fundraiser in Atlanta earlier this month, President Obama is
reported to have said: "It gets you a little nervous about what is happening
to global temperatures. When it is 75 degrees in Chicago in the beginning
of March, you start thinking. On the other hand, I really have enjoyed nice
weather."
What is happening to global temperatures in reality? The answer is: almost
nothing ... 阅读全帖 |
|
l****z 发帖数: 29846 | 5 "Based on our forecasting model, it becomes clear that the president is in
electoral trouble"
by Rick aka Mr. Brutally Honest
Let's all hope the University of Colorado streak continues:
A University of Colorado analysis of state-by-state factors leading to
the Electoral College selection of every U.S. president since 1980 forecasts
that the 2012 winner will be Mitt Romney.
The key is the economy, say political science professors Kenneth Bickers
of CU-Boulder and Michael Berry of CU Denv... 阅读全帖 |
|
r********o 发帖数: 1423 | 6 Price
Despite the fact that final pricing has not been announced, Tesla has been
accepting $5,000 deposits for reservations since it was first unveiled.
Specs and price for the Tesla Model X are expected to roughly align with
those of the Model S.
Prices for the Model S sedan include a 60-kilowatt-hour version for $69,900
and an 85-kilowatt-hour model for $79,900. Many buyers are eligible for a $7
,500 federal tax credit
http://www.plugincars.com/tesla-model-x |
|
t*****g 发帖数: 49 | 7 提車先後跟排隊先後無關, 完全看Tesla高興......
最好的例子就是Model X... 我有同學Model X發佈當天就定了一部低配的Model X, 結
果等了三年了, 他同事大概一年前才訂車的都比他先拿到.... 為啥呢? 因為Tesla為了
賺錢, 宣布目前只做高配的Model X, 低配的Model X到2017年才考慮生產...... |
|
f******t 发帖数: 7283 | 8 绝大部分豪华轿车换代的时候,新闻稿里必然都会有一个章节特别讲车身结构,里面肯
定都会提抗扭刚度;假如优势明显的话,厂方甚至会说具体提升了百分之多少。下面是
W222 S class发布时,官方新闻稿里的一个章节:
Body: maximum stability and high-quality lightweight design
A high level of crash safety, outstanding rigidity for excellent handling
with extremely low levels of noise and vibration. These were the aims when
developing the bodyshell for the new S-Class
– a third-generation aluminium hybrid bodyshell.
– the torsional stiffness in relation to weight and vehicle size – has
been improved by 50... 阅读全帖 |
|
|
c*****t 发帖数: 520 | 9 非常感谢指正,问的确实是应该叫model number,而不是series number 。
在我的印象中,确实有些model number相同的laptop,配置也有区别。所以我才问如何
当场从model number判断性价比这个问题。比方说我今天看到Acer Aspire AS5742Z-
4601,记得和版上近期的某个Acer Aspire AS5742Z 后接某个4 字头的laltop(4685)
差不多价。我只能上网去查印象中的那个4 字头laptop的配置和眼前的4601相比,到底
哪个更高档了?
Serial number对每一台电脑是unique的。 是电脑做好以后create的,里面含有电脑的
一切信息。 而你列举的是电脑的model number, 是一类电脑的型号。 同一model
number的电脑配置基本上是一样的, 偶尔有例外。 |
|
k********a 发帖数: 7225 | 10 我不是做equity research的,但是也要做modeling。
不止输入数据。有时候会combine若干个model,有时候要在model里加些变量,有时候
对现有的model做些修改,etc。
从这个job description看来,似乎不需要你build up a model from scatch.
Good luck! |
|
s****e 发帖数: 1180 | 11 【 以下文字转载自 Statistics 讨论区 】
发信人: sheide (shei), 信区: Statistics
标 题: statistical score modeling, and scoring models 是什么东西?
发信站: BBS 未名空间站 (Fri Jul 1 18:28:18 2011, 美东)
statistical score modeling, and scoring models 是什么东西?有一个职位,要面
这个东西,我google了一下没结果,哪位大大给解释一下,很感谢。 |
|
m****s 发帖数: 18160 | 12 【 以下文字转载自 Actuary 讨论区 】
发信人: kate1998199 (kate1998), 信区: Actuary
标 题: P&C predictive modeling senior analyst or manager job in San Francisco
关键字: P&C predictive modeling job H1B
发信站: BBS 未名空间站 (Tue Jan 22 22:47:27 2013, 美东)
P&C predictive modeling position in San Francisco, senior analyst or manager
, depending on experience. Company sponsors H1B and EB2 is also possible if
qualified.
Requirement:
* 3+ years P&C predictive modeling experience
* Proficient in modeling tool such as Emblem and SAS... 阅读全帖 |
|
t*********p 发帖数: 44 | 13 https://recruiting.ultipro.com/PAC1009/JobBoard/37e95049-80e2-1
f826b780e4d6/Opportunity/OpportunityDetail?opportunityId=f704b3aa-41a1-4f73-
a726-a331cf574b24
RESPONSIBILITIES:
• Obtain and conduct data analysis required for stress testing model
development
• Developing and executing primary and benchmark DFAST stress loss
models for credit risk and PPNR
• Perform all required tests and measures of developed models (e.g.,
sensitivity, accuracy, volatility)
• Deliver compr... 阅读全帖 |
|
s******y 发帖数: 71 | 14 可惜model home的location都不尽理想。
我看中的model home离小区的入口比较近。 而且在corner。 房子是一个小court最外
面的一个,所以corner的一边还
行。 另一边靠小区的街道, 来来往往的的车和行人都能看见你的后院。
唉, model home的装修真是没说的。 而且现在的市场估计能砍下的价钱也不少。 可
是房子的location更重要。
应该是好的选择。 Model home 的
upgrades 很多。Most models are designed by professional interiour designers-
very pretty. 可惜,我买不起。 |
|
w*********y 发帖数: 7895 | 15 今天和LG抱怨说,自己觉的不行,很烦躁,心情不好。。。就问LG,如果他评价
我的能力的话,从1 到 10,我在那里。。LG说8. 我就问,怎么可以提高到9.
他说,很难,因为我没有一个好的role model. 说人提高自己,都是需要一个
role model, 在不断的尝试失败后,通过ROLE MODEL reflect自己,慢慢的
把细小的weakness修正掉。。。所以和聪明人混,自己也会更聪明,和笨蛋
在一起,就会变的脑残。。。
所以我就想,这个是不是和父母是什么样的ROLE MODEL有关呢。如果父母
比较粗心,不讲究细节的,孩子因为没有得到这样的EXPOSURE,是不是也
容易粗心,不讲究细节。。。因为没有ROLE MODEL给孩子表现如果讲究
细节。。。。 |
|
F****s 发帖数: 3761 | 16 所以Model 3不该存在,抢了自己生意还倒贴钱。最麻烦的是,本来Model S十几万起,
现在一群穷人买了几万的Model 3觉得这么便宜就高大上了,殊不知买Model S的人不屑
与买Model 3的为伍,好好的牌子给砸了。
只要Tesla上调车价,让买提嘴不干净的低俗人士买不起Tesla任何型号的车,那么销售
额就慢慢上去了,利润率也提高了,TSLA股价立刻稳住。 |
|
f*********e 发帖数: 8453 | 17 $55K的车就觉得高大上就跟买个果机或者宝马3系觉得高大上一个道理。你觉得买果机
或者3系觉得高大上的人存在吗?如果存在影响果子或者别摸我品牌了吗?
: 所以Model 3不该存在,抢了自己生意还倒贴钱。最麻烦的是,本来Model S十几
万起,
: 现在一群穷人买了几万的Model 3觉得这么便宜就高大上了,殊不知买Model S的
人不屑
: 与买Model 3的为伍,好好的牌子给砸了。
: 只要Tesla上调车价,让买提嘴不干净的低俗人士买不起Tesla任何型号的车,那
么销售
: 额就慢慢上去了,利润率也提高了,TSLA股价立刻稳住。
|
|
d***a 发帖数: 13752 | 18 嗯?Charger SRT是Charger的增强版。Charger和Model 3对应,Charger SRT就应该
和Model 3性能版对比。Model 3性能版的悬挂,刹车和轮胎都升级过。
如果你拿Model 3性能版和Charger普通版来比,那就要秒杀了。Charger普通版比不
过Model 3普通版。 |
|
|
c******a 发帖数: 4400 | 20 Look at these two!!
2017 Tesla Model S P100D vs. 2017 Audi RS7 Performance
Model S beats RS7 start to finish, with 1.4g acceleration!
Model 3 Performance has the same 0-60 mph under 3.5s, which means it could
perform similarly to this Model S and beating the $120k Audi RS7!
19:28 of this video.
Model S beats Porche at launch! |
|
发帖数: 1 | 21 你跟他讨论facts他就来跟你说技术,用几个0-60加速来证明自己多懂技术。倒是发点
有技术含量的干货帖给大伙看看tesla到底有多牛逼啊,楼主还发了个用户体验贴
事实:tesla搞出来的race model GT model S 在2018年刚上赛道一圈半就下来了。
另外一个事实,Chris Harris Drives | Top Gear 搞了个model s P100D和
Porsche911R的 drag race,tesla先要在触摸屏上点N(N<10)次设置max battery
output,然后等十分钟heating(可以随时出发的911r,无奈要等model s heating,
真尴尬),然后两人换车继续比赛,这时model s需要溜一会车来给电池降温,同时
911r在坐着等,lol。这还是直线加速,如果在赛道上不停减速加速,没法让电池冷却
,估计把自己烤熟了吧。
事实:那人连tesla毛都没摸到,就来这胡搅蛮缠。
真想知道这人知不知道自己的X3是几缸的。 |
|
f******t 发帖数: 7283 | 22 发信人: felixcat (Cirrus Logic), 信区: Automobile
标 题: 早就说过Model 3结构设计有重大缺陷
发信站: BBS 未名空间站 (Sat Sep 29 02:30:58 2018, 美东)
其实Tesla目前三款车型都有这个本质性的结构问题:由于设计技术水平低下、没经验
、用料差、省成本(这些原因其中某些,或者是全部一起的组合),竟然采用只有人力
三轮车上面才会看到的skateboard架构。
也就是车底电池盒子就是主要的承力/抗结构性扭动的部件。但你看这个扁扁的盒子能
有多大抗扭的能力?四条边而已,为了电池模块容易拆卸,中间几条金属条是和电池连
在一起的,只能用螺丝拧到盒子上去。这种二维结构怎么可能有效抵抗对角线的应力呢
?(也就是说,四个车轮对角线位置被路面坑洼冲击的时候)
由于model 3卖得多,所以用户抱怨的case更容易积累出来。相当一部分model 3,里程
并不多,那个玻璃车顶就因为车架经常性大幅度形变,应力拉扯已经产生了严重裂痕。
这个毛病各个model 3论坛一搜一大把,比如说这个:
https://model3owne... 阅读全帖 |
|
b****d 发帖数: 35 | 23 【 以下文字转载自 JobHunting 讨论区 】
发信人: budbud (I am bored), 信区: JobHunting
标 题: job opening in northern NJ, key word: statistical modeling, SAS, commerial lending
发信站: BBS 未名空间站 (Wed Oct 16 20:26:11 2013, 美东)
Urgent Hire, expect to start ASAP.有兴趣请站内联系。
AVP, Quantitative Analyst (Credit Scoring Models)
This is a position within the Quantitative Development & Analysis Group.
The incumbent is expected to design, develop, implement and maintain various
credit scores and their applications in credit decisi... 阅读全帖 |
|
f*******2 发帖数: 341 | 24 Qualcomm misses revenue estimates, defends licensing model
http://finance.yahoo.com/news/qualcomm-misses-revenue-estimates-defends-005140993.html
Reuters ReutersJanuary 25, 2017Comment
(Adds analyst comments)
By Narottam Medhora and Stephen Nellis
Jan 25 (Reuters) - Qualcomm Inc reported a lower-than-expected 3.9 percent
rise in quarterly revenue on Wednesday, and defended its licensing model in
the face of multiple legal challenges over its alleged "anticompetitive"
tactics.
The company's sh... 阅读全帖 |
|
s****e 发帖数: 1180 | 25 【 以下文字转载自 Statistics 讨论区 】
发信人: sheide (shei), 信区: Statistics
标 题: statistical score modeling, and scoring models 是什么东西?
发信站: BBS 未名空间站 (Fri Jul 1 18:28:18 2011, 美东)
statistical score modeling, and scoring models 是什么东西?有一个职位,要面
这个东西,我google了一下没结果,哪位大大给解释一下,很感谢。 |
|
n**a 发帖数: 12 | 26 Hello,
Amazon.com is looking for experienced Research Scientist with Machine
Learning, Modelling, Analytics background. Please send your resumes to
n******[email protected]
Many positions open, location- Seattle, WA
Job description: Research Scientist
Product Ads is breaking fresh ground, pioneering a crucial platform for
future Amazon growth. This is high-profile business unit, with support and
interest from all parts of Amazon and top management. We are entrepreneurial
and have a basis for action w... 阅读全帖 |
|
x*****e 发帖数: 233 | 27 看了很多期的America's next top model. 发现其实真正好看的model没有几个。而且
越对称越美的脸越容易被淘汰,用她们的话说,就是你虽然美,但是你没有特点,也没
有什么可造性。the beauty is not the only thing to make modeling。对女模来说
,越冷艳越中性长的越奇怪,脸部线条越明显那么你出位的机会就越多。的确是这样线
条越明显,那么可造性就越大,可以有各种各样的makeups,可以表达各种各样的theme。
里面很多选手,都不好看,很一般,但是每次照片出来,效果都惊人的好,惊人的漂亮
,一是因为摄影师技术好,而是脸部线条明显,那么一定可以拍出好看的照片。越到最
后其实模特的素质条件显得不那么重要了,因为其实她们都差不多,拥有完美的脸部线
条,身材,身高。所以passion就变得那么重要,有时候我觉得美国梦其实就是你对一
件事的passion。到最最后,性格也变得很重要,性格好的就容易得胜。
从America’s next top model。里面有一个选手说,Simple is all you need。觉得
说的很... 阅读全帖 |
|
s*****o 发帖数: 48 | 28 CFP: CVIU Special Issue on Shape Modeling in Medical Image Analysis
Aims and Scope: Exploring and using anatomical shape properties plays an
increasingly important role in many medical imaging and image processing
applications. For examples, in medical image segmentation, prior knowledge
on the shape of interest has been used to achieve more accurate segmentation
, e.g. in case of ambiguities owing to limited image quality. In
neuroscience, the shape of certain brain structures, such as hippoca... 阅读全帖 |
|
w*******7 发帖数: 188 | 29 You can use "Entity Framework" to create a model from a database, WPF and
Window Form can consume this model. I am not sure about Webform (I think it
could).
But for MVC, the reverse engineering model is used ObjectSet and Objectquery
, Is there any way to implement the model directly in MVC as Data Domain?
Or you have to use code first to build the data model and set connectting
string to direct connect to database? |
|
i*******e 发帖数: 242 | 30 不知道我的思路对不对
问题是这样:我的dataset里的response有两个,一个R1是从0-1的分布;另一个R2是切
割R1分了4个band:0 ->0; (0-0.3] -> 1; (0.3-0.8] -> 2; (0.8 - 1] -> 3. R1的分
布不太好,“0”很多,其他的是left-skewed分布。最终目标是model R2.
直接model R1, mean error 大概0.2, 但是model R2时 试了很多classifier 都在0.
6以上。我估计因为prediction的时候如果是0.1,虽然很接近实际值0,但是在R2分类
还是算1,导致error变大。
我琢磨能不能先用modelA判断是零还是其他,再用一次model判断是1,2,or3.
1. 这种用多次model分类的有好的例子借鉴吗
2. 不知道有没有其他的方法。
多谢! |
|
w********h 发帖数: 12367 | 31 如果你现在的文章不能很快写出来得话,
先让老板跟你一块发表这个review占领这个model的发现权。
对你后面发表文章没啥影响,
因为这个review的重点肯定不是这个model吧。
你后面的实验出来之后反而确证这个model的用处,一样重要。
没有这个model在里面,你可能就不是这片综述的作者了,
而且跟老板关系最重要,就看他承认你的贡献并且表扬你我也觉得他是个reasonable的
人。
当然,有顾虑也是可以理解的,你可以试着问问老板会不会对你后面的文章有影响。
他肯定会说我前面的话。还是记住这点:你的成果被老板scooped你应该感到honored,
因为他会在其他方面给出补偿和在推荐信上大说特说。
更何况根本不是那回事,你的名字不是在么?
有的不良老板在学生的工作启发下甚至学生的努力总结下写好一片综述,
这片综述的主要思想都是这个学生发展的,而且是另外一篇research article需要用的,
然后提前发了,不带学生的名字,仿佛学生就一labor,
这种事情也很多的。你应该庆幸你没遇到这样的trash。
上present这
了进展,就跟老
model
今天老板发来草 |
|
a*****x 发帖数: 901 | 32 附件是一个简单的conventional model和paddle model的示意图。
conventional model基本上就是一根voltage sensor棍棍移动,也就是我之前说的描述
voltage sensor的大致正确的模型。
paddle model认为移动的不止S4,而且S4的移动包括了旋转,和两端带电helix之间位
置的变化。看图很容易明白。
paddle model我印象中由MacKinnon最早根据一些很间接的证据对钾离子通道提出,后
来有不少点突变的证据。但是就我的印象,没有smoking gun。也许后来有了,希望有
人介绍一下。
voltage |
|
c******d 发帖数: 1150 | 33 【 以下文字转载自 Statistics 讨论区 】
发信人: caldavid (认识), 信区: Statistics
标 题: 有没有什么软件可以实现用多种model来fit data的?
发信站: BBS 未名空间站 (Thu Feb 12 22:38:37 2015, 美东)
matlab我知道的能一个model一个model试,但是需要试的model太多了,想问问有没有
什么软件能够只要输入自变量和因变量就能使用不同model来模拟,算误差的。。
谢谢 |
|
c******d 发帖数: 1150 | 34 【 以下文字转载自 Statistics 讨论区 】
发信人: caldavid (认识), 信区: Statistics
标 题: 有没有什么软件可以实现用多种model来fit data的?
发信站: BBS 未名空间站 (Thu Feb 12 22:38:37 2015, 美东)
matlab我知道的能一个model一个model试,但是需要试的model太多了,想问问有没有
什么软件能够只要输入自变量和因变量就能使用不同model来模拟,算误差的。。
谢谢 |
|
N********m 发帖数: 107 | 35 【 以下文字转载自 JobHunting 讨论区 】
发信人: Nightstorm (Swordblade,Nightstormy, Icyriver), 信区: JobHunting
标 题: Modeling/Scoring/Analysis Analyst, manager and senior manager positions
发信站: BBS 未名空间站 (Wed Apr 8 12:16:09 2015, 美东)
Consumer modeling unit in a major bank is under big expansion. There are
many positons for entry and experienced level canidates. They are hiring
Modeling/Scoring/Analysis Analyst, manager and senior managers.
Requirements: MS/PH.D. in Stats, econ or math.
Experience in using econo... 阅读全帖 |
|
s*****o 发帖数: 48 | 36 CFP: CVIU Special Issue on Shape Modeling in Medical Image Analysis
Aims and Scope: Exploring and using anatomical shape properties plays an
increasingly important role in many medical imaging and image processing
applications. For examples, in medical image segmentation, prior knowledge
on the shape of interest has been used to achieve more accurate segmentation
, e.g. in case of ambiguities owing to limited image quality. In
neuroscience, the shape of certain brain structures, such as hippoca... 阅读全帖 |
|
s*****3 发帖数: 1673 | 37 LZ在Malta还是 Albany ? PhD? 那个modeling是Process, device modeling还是偏电
路一点的compact model之类的
我觉得NY那边就IBM,GF吧,其他的半导体公司大部分都在西部,AZ,CA, TX之类的。包
括IC的公司这边也很多。做Device research 也没几个了。面的确很窄,但是基本上就
在那几家工作了,不会有大的变动,做的东西肯定是前沿。modeling基本上IDM都会有
,fabless也需要会process和modeling的人。看你喜欢那个了?如果我是LZ我就去CA了
。 |
|
s*****o 发帖数: 48 | 38 CFP: CVIU Special Issue on Shape Modeling in Medical Image Analysis
Aims and Scope: Exploring and using anatomical shape properties plays an
increasingly important role in many medical imaging and image processing
applications. For examples, in medical image segmentation, prior knowledge
on the shape of interest has been used to achieve more accurate segmentation
, e.g. in case of ambiguities owing to limited image quality. In
neuroscience, the shape of certain brain structures, such as hippoca... 阅读全帖 |
|
K******o 发帖数: 296 | 39 别的组做了model,送到Model Validation来检查,这帮人就开始挑错挑刺,然后让做
model的人回去修改,提供证据.反正就是给做model的人找各种麻烦,美其名曰:Model
Risk Management
其实就是BASELII给创造的工作机会而已。 |
|
m*****a 发帖数: 4 | 40 A leading firm has 12+ month contract opportunities with long-term potential
in downtown NYC for a Quantitative Analyst.
Key Responsibilities:
Support the execution of data-driven analytical opportunities that
significantly improve profitability, financial and risk management controls.
This position requires data analysis, cleansing and reporting to support
predictive modeling. Responsibilities include large data manipulation,
running various statistical models and strategy implementation.
Ad... 阅读全帖 |
|
n****e 发帖数: 2401 | 41 金融学多了脑子都一滩浆糊。本来model就都不太对,大概是那么回事就行了。然后用
model来fit market data,反着解出个imp vol。然后大家又拿这个来解释market。
不管是lognormal还是normal model,都是用来fit market data。At the end of the
day, all models should get the same market price, then all models have
different imp vol. It is not about which is right which is wrong. |
|
c********s 发帖数: 101 | 42 Interesting my post provoked a discussion around MV. Here is another of my
posts. When someone working in MV ask about difficulties in work, I replied
this. Just add to the discussion for fun. Someday I will dig a "Keng" about
MV, so many things to say, feels like they want to burst out of me...Maybe
when my boss is on vacation next week. BTW, we used out every single day of
our vacation. Last time I checked, the bank is still functioning...guess we
are not as important/successful as manager Liu... 阅读全帖 |
|
r**a 发帖数: 536 | 43 This is a good post. I totally agree with you. Actually in my previous posts
I should emphasize that the MV team is normally doing the supportive thing.
As a MV guy you are normally far away from the decision maker. So, even you
found a better model than the submitted one, it is not you who will decide
which model should be used. But as part of the job duty, you need to make
some comments on what you found and let the decision maker choose. For me I
believe that ideally a good model validator sh... 阅读全帖 |
|
t********t 发帖数: 1264 | 44 Some MV head is actually very touch and tactical on this, especially when he
can see the point that most of the models are already well developed and
organized in the library. Less and less complex flow trades drive no demand
for super-sexy model anymore. When MV can be turned into a model centre for
the whole bank, he will fundamentally change the bank's corporate structure,
not to mention his own rank.
Here I am talking the FO models, not the CCR models. Well CCR is hot and be
the first priori... 阅读全帖 |
|
s*****n 发帖数: 2174 | 45 你的一些基本概念有些混淆.
a good (fit) model 和 prediction accuracy 没什么直接关系.
model只是用来 fit observed data. 然后以一定的标准(比如最小平方和)来评价好坏.
至于prediction, 这严重依赖于你做predict的时候的assumption和你data本身的性质
. 你的data本身noise term就很大, 也许无论如何你都无法精确predict. 你再怎么找
model也没用.
很小的R^2并不一定说明 model 不好, 或者是存在更好的 model. |
|
w**********y 发帖数: 1691 | 46 The easiest way might be AIC, BIC..
Further, Bayes Factor..But I don't think you can do that directly in R and
SAS...
Further, en, that is in the research of model selection.
I just realized, what you mean by HIERARCHICAL? they are hierarchical models?
Or do you mean they are nested?== Model A <=Model B?
If so, take suggestion from lapringles ..Fit each model by SAS, then get the
score of -2logL |
|
B****n 发帖数: 11290 | 47 causal model裡面的箭頭只能是單向的 因為causality是有先後的關係
structural equation model則無此限制 所以應該不能說structural equation model
一定是causal model
VAR |
|
c*****k 发帖数: 42 | 48 请教版上的统计Ph.D.们一个model selection problem.
问题:两种不同方法estimate出不同的结果。这两个模型都有各自的优缺点。现在想把
这两个模型给合并起来。
方法:
(1)简单的平均两个模型的estimation结果
(2)Bayesian Model Averaging?
(3)Model combination models?
(4)其他?
请给出一些文献,关键字,例子。
多谢! |
|
K***a 发帖数: 72 | 49 我是用SAS Eminer做model,不是很精通response model,但知道大概过程,
想知道在做法上或结果上,response model和look-alike model有什么不同? |
|
s*********e 发帖数: 1051 | 50 response model is predictive with the purpose of targeting.
look-alike model is descriptive with the purpose of profiling.
response models, for example, are built directly off promotional files and
corresponding responder information.
look-alike models, on the other hand, are not based on promotional
information. Instead, they evaluate how similar in appearance each
prospect is to a direct marketer's customers, in an indirect attempt to
capture the dynamics of response behavior. |
|