l****z 发帖数: 29846 | 1 Tesla to unveil mainstream, cheaper model in 2015
By Sarah Drake, contributor, Silicon Valley Business Journal
CEO Elon Musk stands with Model S wheels at Tesla's plant in Fremont, Calif.
Musk's company plans to unveil a less-expensive Tesla version in 2015.
Electric car maker Tesla Motors will tap into the mainstream car market with
a vehicle that could sell in the $40,000 range, and could become crucial to
its future growth.
The Palo Alto, Calif.-based company run by Elon Musk could introduce ... 阅读全帖 |
|
f*******e 发帖数: 5277 | 2 怎么跟你说呢……背景知识太少我解释起来还挺费劲。
铺垫的意思是建超冲网络,建门店,搞研发,以及最重要的,盖那个世界上最大的电池
厂。这些东西对于model s和model x来讲都是过剩的。比如超充网络,除了加州的一部
分超充点以外,大部分超充点的利用率都不到30%吧,但如果要等到今年才建,model 3
就跑不了长途了,所以只能提前布局。Gigafactory更是如此,不做3的话,根本没必要
去盖这个5 billion的工厂。但你为了配合3的量产,必须提前几年砸钱去盖。而且这些
铺垫越多,每个季度以折损的方式计入OpEx进入财表的数字就越大。
再说产能没信心,特斯拉原计划是2020年的时候产率达到每年五十万辆。结果去年
model 3第一次发布的时候预订爆棚了,决定把这个量产计划提前两年。这确实很困难
,没人否定这一点,我自己怀疑特斯拉明年年底的时候能否达到每周一千辆的水平。但
Musk没那么有信心的是ramp up速度。他去年就说了,计划2017年7月31日能开始量产,
并说这是个inspirational deadline,主要是鞭策自己员工以及所有供应商,不指望能
真的在7月3... 阅读全帖 |
|
T*******y 发帖数: 560 | 3 北京时间8月21日晚间消息,特斯拉电动汽车的“Ludicrous模式“(荒唐模式/狂暴模式
)的确挺荒唐的,近日在一次竞赛中竟然战胜了兰博基尼专业跑车。
当前,Model S P100D是特斯拉的最强悍车型,在”Ludicrous“模式下,从0到60
英里/小时(约合96公里/小时)加速仅需2.5秒钟。上周末,YouTube用户DragTimes上传
了一段视频,再次证明了这一点。
视频中,一辆特斯拉Model X P100D跨界SUV与一辆价值53万美元的兰博基尼艾文塔
多(Lamborghini Aventador)旗舰跑车对决。而结果是:特斯拉领先兰博基尼0.05秒。
重量方面,Model X P100D重5400磅(约合2449公斤),较兰博基尼重1200磅(约合
544公斤)。但是,在“荒唐模式”下,Model X P100D不仅战胜了兰博基尼,而且还创
下了速度最快SUV的记录(400米)。
在1/4英里(约合400米)的距离赛中,特斯拉用时11.418秒,领先兰博基尼0.05秒,
速度为每小时117.95英里(约合每小时190千米)。比赛时,这台Model X P100D... 阅读全帖 |
|
|
c********n 发帖数: 152 | 5 Statistical Modeler / Statistician needed for our firm. Lead development of
predictive modeling and analytics projects using data mining techniques (
including linear and logistic regression, neural networks, decision trees,
etc). Project and measure the effectiveness of modeling efforts and
marketing actions such as targeting, prospecting; Build predictive models
and marketing analytics.
We are looking for a career-minded, bright, motivated individual to be part
of our analytics team. Knowledg... 阅读全帖 |
|
t*********p 发帖数: 44 | 6 【 以下文字转载自 JobMarket 讨论区 】
发信人: thebodyshop (body...), 信区: JobMarket
标 题: [招人]Quantitative Modeler
发信站: BBS 未名空间站 (Tue Apr 19 18:04:40 2016, 美东)
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 ... 阅读全帖 |
|
c**********a 发帖数: 92 | 7 The Company:
Faraday Future (FF) is a California-based mobility company, leveraging the
latest technologies and world’s best talent to realize exciting new
possibilities in sustainable transportation. We’re producing user-centric,
technology-first vehicles to establish new paradigms in human-vehicle
interaction. We’re not just seeking to change how our cars work – we’re
seeking to change the way we drive. At FF, we’re creating something new,
something connected, and something with a global impac... 阅读全帖 |
|
h*******d 发帖数: 8 | 8 Center of Atmospheric Science
University of Illinois at Urbana-Champaign
This position is aimed to improve the capabilities of existing atmospheric
dispersion models to stimulate local and regional pollen and seed dispersion
, deposition, and outcrossing. Models will be converted to an online tool so
stakeholders can easily access and use.
Duties and Responsibilities: Measure and model plants atmospheric gene flow
, including pollen and seed release, dispersion, deposition and outcrossing.
An... 阅读全帖 |
|
|
f*******e 发帖数: 5277 | 9 我来说一些相关的我个人的看法和计算
1. 35k给那个配置的话我觉得是相当合理,即便没有7500的tax credit。从发布后一天
预订又翻了一番来看,至少后面那些人是不在乎这7500的,否则在各种新闻提示phase
out的情况下,如果没有7500的tax credit不买的话,就没必要预订了。我在网上看的
几百个帖子和排队时问的十来号人的情况看,大家普遍觉得能拿不到7500的话就少上几
个option,不会不买。
2. 现在预订的人我觉得,要么是R/S/X车主给孩子买的,要么是试驾过S后再也不想开
油车但又买不起S/X的(这部分应该算粉丝)。由于预订限制每人只能订两辆,而且车
最快最快一年半后才能拿到,所以黄牛出现的可能性不大。只要真正交车的时候没有大
的变动,后面不说,前250k的预订转换成实际订单的比例应该很高。
3. 成本是最大的问题之一,所以我花时间算过一次,大致如下
油车的动力系统平均成本(包括SUV和皮卡)大约是$7.5k(http://insideevs.com/electric-cars-to-cost-same-as-or-less-than-ice-withi... 阅读全帖 |
|
|
a****o 发帖数: 6612 | 11 Elon Musk一度扬言计划model 3的生产达到每周10000辆,年产50万辆,单一型号已经
超过toyota camry(2017 年美国销售38万7千),当然model 3会是全球发售。虽然model
3的价格定位是宝马奔驰的轿车系列,但是其生产销售目标可不是。
宝马的3系和4系全球总共在2017年卖了99万辆。
Camry的销量减少,但是rav4的销量在增加。貌似消费者倾向于开“大”的车。从这点
看,并不利于model 3的销售,要等model y出来---如果能撑到model Y的话。 |
|
发帖数: 1 | 12
. ”
Their SEC filing (Form 10-K) said M3 was for mass market:
http://www.sec.gov/Archives/edgar/data/1318605/000156459018002956/tsla-10k_20171231.htm
Page 1:
"We commenced deliveries of Model 3, a lower priced sedan designed for the
mass market, in July 2017 and continue to ramp its production."
Page 5:
"Moreover, we have designed Model 3 with a mix of materials to be
lightweight and safe while also increasing cost-effectiveness for this mass-
market vehicle."
Page 17:
"Our future growth and su... 阅读全帖 |
|
a****o 发帖数: 6612 | 13 德国公司把model 3拆开,估计材料成本$18,000,人工成本$10,000,总成本$28,000。
虽然不清楚具体的成本分析,但觉得如果这个数据可信度高的话,那么我觉得model 3
很难做到盈利。
可以做比较是苹果的iphone X,64G款的成本是$370, (https://www.fool.com/
investing/2017/11/13/apples-iphone-x-reportedly-costs-370-to-build.aspx)
售价是$999。
类比下来,model 3要卖到 75600,才能达到类似苹果的盈利。考虑到iphone X在人力
成本低的国家地区生产,生产量大(虽然单价低,但是生产量是千万级别,并且没有所
谓的production hell), model 3的售价需要更高很多,才能到达类似苹果的持续性的
盈利。
当然,德国公司的发现中亮点也有,就是钴在阳极材料总的含量(不清楚具体单位,可
能是锂镍钴的摩尔比例)只有2.8%。 通常来说,镍的含量高,那么电池的容量高,而
钴的含量高,那么稳定性会高。如果这个规律仍然正确的话,那么与18650相比... 阅读全帖 |
|
c******e 发帖数: 82 | 14 Again, carry on you Tesla bashers, pro or non-pro.
Received my model 3 on Tuesday. The car is a game changer, there’s no doubt
. It’s the future and nothing I’ve ever driven with a gas engine can
compare + at this point I’ve been converted like probably the rest here who
have received their Model 3’s. EV’s from here on out for me.
I won’t lie, some of the reviews, negative media attention had me nervous
leading up to getting my car. I was nervous about shoddy workmanship,
misaligned panels, chea... 阅读全帖 |
|
f**********g 发帖数: 4709 | 15 难道自动驾驶不是最大卖点和最重要的功能吗?东拉西扯说了一堆到了重点了来个不说
了,
就跟女的买了双名牌鞋得瑟一堆这里好那里好,那么能不能穿着走路呢我就不说了,
观众表示不明白,您还是详细说说吧
: 事实,想科普一下:
: 虽然 model 3 车子小,比model S/X 便宜,但是电动车核心技术和很多设计比
现在 的
: modelS/X 领先一代:
: 1. 电池包2170的能量密度超过 model s 18650 30%。 另外, 成本大大降低,
因为钴%
: 已经低于5%,未来5个月-一年要实现完全没有钴。 下一代model SX 很快采用
新技术
: , 马上要推出130 度电电池包。
: 2。逆变器全部采用sic 技术,虽成本较高(每个die 本身成本增加),但是因
为效率
: 提高~1.2%,导热性能极大提高,耐高温,电池里程提高的电池成本节省可以抵
消逆变
: 器成本增加。 实话实说,model S 本身逆变器技术就已经行业大幅领先,主
要是自
: 己封装技术远远超过其他公司,包括英飞凌... 阅读全帖 |
|
t*********p 发帖数: 44 | 16 【 以下文字转载自 JobMarket 讨论区 】
发信人: thebodyshop (body...), 信区: JobMarket
标 题: [招人]Quantitative Modeler
发信站: BBS 未名空间站 (Tue Apr 19 18:04:40 2016, 美东)
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 ... 阅读全帖 |
|
m********l 发帖数: 6748 | 17 1个半小时哦,就是需要交$299的入场费。
1:00pm - 2:00pm:
Instructor Manny Cabo will review some of his best work with nude and
implied nude models and show you lighting techniques to capture stunningly
beautiful body highlights and tasteful images celebrating the female form.
2:00pm - 3:30pm: OPEN MODEL SHOOT - Attendees can photograph an assortment
of beautiful models. Some models will be fully nude while others will be
posing in lingerie. Each station will be lit with continuous light.
3:30pm - 4:00pm:
Using im... 阅读全帖 |
|
w*****g 发帖数: 16352 | 18 最SB的不是这个。
三万五的充一次电只能跑200迈。要想达到S型的300迈一充,要再加九千买附加电池。
但这样车重增加,操控就差了。
直接说是猴版model S不得了。
[在 yeson (快乐着并幸福着...) 的大作中提到:]
:Musk真是人如其名,皮厚的程度在大忽悠群体中也是厚的无以复加。
:信源:华尔街见闻|编辑:2017-07-30
:对于翘首期盼已久的特斯拉粉丝们来说,特斯拉首款平价电动汽车Model 3的交付仪式
:可谓惊喜和失望并存。
:最大的惊喜在于,作为入门款车型,Model 3在性能上并没有打折,但这款“平民”特
:斯拉的定价则遭到诸多吐槽。吸引眼球的35000美元只能买到基础黑色款,如果你想换
:个颜色,或加点配件,价格就蹭蹭地跳涨。更令人失望的是,昨日交付的30台都是高
配车型,特斯拉也表示初期将集中生产高配版,35000美元的基础版车型要在今年秋季才
:能交付。
:Model 3被寄予厚望,是马克斯完成其“远大规划”的最后一步,Model 3成功与否很
大程度上将决定特斯拉,甚至电动车市场的未来发展,如果Model 3能成功打入主流市场
:,或将开启电动... 阅读全帖 |
|
d*****u 发帖数: 17243 | 19 graphical model:基本上能用图表征的model都算
language model:应该是一个应用。
bayesian network: 只是GM的一种,基于变量之间的条件概率
HMM:也可以看成是graphic model的一种,是一种树形的Bayes net
N-gram:就是N阶的Markov chain.其实也可以看成非树形的Bayes net。
Markov Model: 是undirected graphical model的一种。
一个变量的概率分布只取决于其neighbors的取值。
类的 |
|
p******m 发帖数: 65 | 20 现在在一个公司做predictive modeling, pricing 部门的。有两个机会可以跳:一个
是做reserving,另一个是换一个公司做senior predictive modeler.
我本身是actuary background,没有统计学位。做了几年的predictive modeling,没有
做过传统的pricing或者reserving.
现在有点纠结,从哪一个机会会对未来发展更好呢?
换reserving吧,好处是可以充实我actuary的background. 不太好的地方就是
reserving感觉上会比较boring。如果我以后还是想回到pricing/predictive modeling
这块,reserving 的experience可能是鸡肋。
做senior predictive modeler, 好处是工资会涨不少,也是我自己喜欢做的东西。不
好的地方就是,我没有统计学位,感觉纯粹做predictive modeling对我来说不是长远
之计。actuary 总需要做过reserving和pricing的。
希望能得到过来人的建议! |
|
b*****1 发帖数: 17 | 21 上次发了一个model organism的帖子,没想到引起那么多关注。我也的确学到很多,非
常感谢各位前辈。
我总结下自己最重要的心得就是:mammaliam work doesn't belong to model
organism work; and for model organism research, if cannot find a faculty
position, then no way in industry.我个人还是很喜欢科研,但是也不会放弃
industry这个糊口的退路。那么就以mammalian work为主,除非一些做model organism
的工作非常特别,或者有价值,或者是special technique-driven
不过这样又引出我的一些问题:
1。mammalian work包括mouse model和直接的mammalian cell line。我觉得做mouse
model那么长周期,工作量大,而cell line,随便一养就可以了。这两者不可同日而语
吧?做mouse那么辛苦,肯定比一个单纯的cell line要高级些吧?
那么假设我 |
|
j***l 发帖数: 100 | 22 我在自己课题中做了一个model,其predictions可以用实验检验。几个月前我在组会上present这
个model,老板会下闲聊时感谢我做的model让此项目变得有趣。其后我的实验上也有了进展,就跟老
板谈可否开始写文章,可老板一直认为还要做其它的实验(我认为那些实验偏题了)。
上周末老板突然来找我说他在写一篇review,正在赶稿,其中有些mathematical model
analysis希望我帮他参考,还有,他考虑我的项目和这个主题相关,就加我为三作。今天老板发来草
稿,我一看,那图,那model idea 可不就是我想在自己paper里写的主要信息么.不知道是我太敏
感,还是他糊涂;我认为这样是被scooped了。老板让我今晚看玩草稿提意见给他,可我真不知该如何
回信。他说了他的文章明天早上就想投出去,我应该是改变不了了。但是我该装作若无其事么? 那我以
后自己的paper怎么办?我猜他是想让我的paper以实验为主,但model是我提出来的,而且这review
发表出来后我再投文章,是不是credit会降低很多,也无法投好杂志呢。
我十分不理解,想寻求真诚意见,不要 |
|
c**********a 发帖数: 92 | 23 The Company:
Faraday Future (FF) is a California-based mobility company, leveraging the
latest technologies and world’s best talent to realize exciting new
possibilities in sustainable transportation. We’re producing user-centric,
technology-first vehicles to establish new paradigms in human-vehicle
interaction. We’re not just seeking to change how our cars work – we’re
seeking to change the way we drive. At FF, we’re creating something new,
something connected, and something with a global impac... 阅读全帖 |
|
l*******G 发帖数: 1191 | 24
I'm doing coastal ocean modeling for a few years, recently focusing on
using ROMS model for estuarine forecasting/nowcasting. ROMS is the leading
model now in US and it's relatively complicated system.
Yet there have been new proposals funded by NOAA/EPA to do regional
environmental forecasting, i.e. couple regional ocean model such as ROMS,
POM or CH3D or FVCOM with regional atmospheric model such as WRF and land
use models such as SWAT etc for an integrated system. A lot of focus i |
|
s********n 发帖数: 154 | 25 The full name for this model is: Model of Acidification of Groundwater in
Catchment. I searched for several days through internet but cannot find it.
If anyone has this model or knows the website to download it, could you
please give me a reply? Thanks in advance!
In addition, I am looking for a model to simulate the watershed
acidification. MAGIC and ILWAS are the most frequently used models. Anyone
knows which one is best, or you have any other better model to recommend? My
watershed is tiny, |
|
N********m 发帖数: 107 | 26 【 以下文字转载自 Economics 讨论区 】
发信人: Nightstorm (Swordblade,Nightstormy, Icyriver), 信区: Economics
标 题: Modeling/Scoring/Analysis Analyst, manager and senior manag (转载)
发信站: BBS 未名空间站 (Wed Apr 8 12:17:45 2015, 美东)
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... 阅读全帖 |
|
N********m 发帖数: 107 | 27 【 以下文字转载自 Economics 讨论区 】
发信人: Nightstorm (Swordblade,Nightstormy, Icyriver), 信区: Economics
标 题: Modeling/Scoring/Analysis Analyst, manager and senior manag (转载)
发信站: BBS 未名空间站 (Wed Apr 8 12:17:45 2015, 美东)
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... 阅读全帖 |
|
c********g 发帖数: 1106 | 28 制药界工作机会比较多的modeling种类:
Discovery: PK/PD modeling, PBPK, mechanism based PK/PD, systems pharmacology
modeling (通常在DMPK)
Development: PK/PD (pharmacometrics, use NONMEM, R, monolix), PBPK (use
Gastroplus, simcyp) (通常是专门的modeling组或和clinical pharmacology一起)
掌握modeling&simulation理论的话,这些东西基本能融会贯通。当然并不是什么都能
一下就会了,但是至少很容易有个大致理解。
PBPK可以作为进入这个行业的敲门砖,进去了有很多学习机会,可以慢慢积累做各种模
型的机会,也可以转向DMPK rep或者clinical pharmacology.
PBPK目前有毒理和药理两个主要的领域。毒理主要是辅助环境危险度评价,有社会价值
,但工作机会少。要了解PBPK,窃以为应该从毒理界的文献开始。Hamner Institute是
计算毒... 阅读全帖 |
|
f*********e 发帖数: 13 | 29 看到一个quantitative modelers的招聘, 里面要求market risk VaR models, 请问实
际工作中market risk VaR都有哪些model?
在john hull 的书上看到计算VAR 的方法 historical simulation, model based
method (linear model, quadratic model), simulation methods。 请问, 实际中
工作中那些方法比较常用?
对VAR的stress test, 实际中是不是主要用simulation 的方法?importance sampling
, Brownian bridge techniques or other, 实际中那些方法比较常用?谢谢! |
|
z****u 发帖数: 185 | 30 SABR is popular in interest rate modeling while Heston is popular in equity
modelling. Many quants like Heston as it is an affine-model with a semi-
analytical solution, and it captures most of the features needed for skew
modeling. On the other hand, SABR does not have the mean-reversion feature
of the vol.
Both models have limited calibration capability. |
|
j******m 发帖数: 70 | 31 请教一下版上的大侠一下:
1. Insurance Company (比如 SunLife or Manulife) 的 Model Validation组都做什
么? 自己感觉好像用各种Statistical Model多一些, 哪位大侠能具体举两,三个例
子吗? 能给几个key word, 小弟就非常感谢了。
2. 和银行的Model Vetting/Validation有很大区别吗? 好像 Insurance Company 的
MV 基本不用 Traditional 的 Derivative Pricing Model. 如果在 Insurance
Company 的 Model Validation 做过一两年, 如果要转银行的Model Vetting或者是
Desk Quant? 是不是很难了? 好不容易从数学转到金融,非常害怕再转错行,希望能
一次选好。 谢谢 |
|
t*********p 发帖数: 44 | 32 【 以下文字转载自 JobMarket 讨论区 】
发信人: thebodyshop (body...), 信区: JobMarket
标 题: [招人]Quantitative Modeler
发信站: BBS 未名空间站 (Tue Apr 19 18:04:40 2016, 美东)
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 ... 阅读全帖 |
|
s*****n 发帖数: 2174 | 33 你问的问题, 本质上是logitic regression model 的
model diagnostics. 这个问题本身就是一个困难的问题.
你要考虑的, 不是给Y分组, 而是给X分组.
分两种情况
(1) Logistic model with replication
同样的X下, 有多个Y(0/1)的观测. 这种情况比较容易.
你可以比较
Y hat = Model(X)
Y empirical = #1/(#1+#0) under X
然后算一下correlation什么的.
(2) Logistic model without replication
同样的X下, 只有一次Y的观察, 或者是0或者是1. 这种
情况下, 必须借助额外的assumption. 常用的就是model
的连续性, 即相似的X意味着相似的P(Y=1)
这时, 需要把所有的observation根据X进行clustering.
然后在每个cluster内, 看成是replication. 进行第一
种情况那样的算empirical rate和predicted rate.
可是如果X的维度较高, 高 |
|
m*****e 发帖数: 13 | 34 Deal All,
I want to ask a stupid question about the null model fit statistics in proc
logistic. I want to get the -2LogL of the null model. I thought there are
two ways to do that.
1) Run the model without specifying any independent variable, i.e.
proc logistic data = mydata descending;model youtcome = /firth;run;
In the output, I found:
Model Convergence Status
Convergence criterion (GCONV=1E-8) satisfied.
-2 Log L = 38.858
2) Run the model by specifying independen |
|
D******n 发帖数: 2836 | 35 combine没啥办法,一般就是voting和average,关键是那些models是怎么产生。
bagging是用不同的bootstrap data去train model,然后平均这些model
boosting就是每步都变化example的weight去train一个model,最后也是加权平均这些m
odels,
你只有两个model,也没啥办法了,BMA,好像实际效果不好,你可以看看。
这些东西都有个统称,叫ensemble method,当然不同field可能有不同叫法。或者叫co
nsensus。
Squares |
|
c*****k 发帖数: 42 | 36 多谢!
如果其中的一个model只有estimation 结果,就是说也不知道具体是怎么做的;但是第
二个model就是利用GLS加上新的数据;那么如果想把两个model的结果combine起来,有
这种方法吗?
因为看您给出的方法,好像都要train之前的model,所以要需要每个model的实现方法
和数据吧? |
|
s**********7 发帖数: 24 | 37 最近在做一个project,有一个数据,是一群人的考试数据,有 ID, binary
response 0/1, 0 是错误回答,1是正确回答, 以及age, gender, years of
education 这样的variable.
Psychology/measurement 里有IRT model, 所以要求用这一堆数据和Rasch
Model (one parameter logistic Model)来估计item difficulty parameter. 这个比
较简单,用HLM,item 是level1, 人的能力是 level 2.item nested within 每个人,
PROC GLIMMIX 就得出来了。
还有一个要求是用 SAS PROC GLIMMIX,用这个binary response 数据和age
, gender, yearsofeducation variable 来model 人的能力. 我有一点没想通,人的能
力是latent trait,在HLM (Multilevel mixed-m... 阅读全帖 |
|
s********8 发帖数: 50 | 38 there are more than 5 variables in the model: A.B.C.D.E.
A. B. are not significant (p>0.05) when they are in the model separately or
in the model togather. But when I put A|B (=A,B,A*B) in the model. Both A(p=
0.01) and B(p=0.02) are significant , but interaction term is not
significant (p=0.08).
what does it mean for modeling ?
do I need put three of them into final model?
Thanks |
|
c******g 发帖数: 63 | 39 如果model是固定的(比如一个linear model,就五个系数要fit),给50个数据(X,y)
,得到fit出的model A;再给50个数据,用这100个数据得到Model B。那么,应该选哪
个model呢?用什么方法来pick model?
我个人觉得,应该拿额外的test data看test error吧。但是如果一共就只有这100个数
据,没有额外data,怎么办?
多谢指教! |
|
p********a 发帖数: 5352 | 40 ☆─────────────────────────────────────☆
zhongdianshi (brb) 于 (Mon Aug 29 09:50:26 2011, 美东) 提到:
OUTCOME: BMI
PREDICTOR: QUESTION1, QUESTION2, QUESTION5, QUESTION6...
所有的PREDICTORS是ORDIANL VARIABLE.
我想分别TEST OUTCOME和每一个PREDICTOR的CORRELATION.
我用了2个方法:
1.
PROC CORR SPEARMAN;
VAR BMI QUESTION1n QUESTION2n...;
RUN;
生成一个CORRELATION TABLE.
2. ANOVA
分别把每个PREDICTOR和BMI放到MODEL里,这一步,我不是很确定.
proc glm data = DATA;
class QUESTION1;
model BMI = QUESTION1;
meansQUESTION1;
run;
quit;
最终,是要建个MIXED MOD... 阅读全帖 |
|
A*******s 发帖数: 3942 | 41 i have the same doubts on some econometric models too. i think it is
something about "model is king" vs. "data is king" question. my 2 cents, in
predictive modeling, "data is king" approaches usually perform better since
it is directly optimized based upon ur purpose. in explanatory modeling, "
model is king" seems more reasonable but it is hard to be verified.
so
sample有可能是biased。为啥不用Heckman呵?
的结果都很相近。难道大多数fancy的econometric model都是用来抵御可有可无的质疑
的吗? |
|
B*H 发帖数: 10 | 42 不好意思,对 mixed model 一知半解.现在想求 full model 中 fixed effects 的 R2.
我的理解是先 run full model,再 run null model, null model 中去掉所有 fixed
effects (intercept only), 保留 random terms (intercept, slope), 另外要让
random effects covariance 与 full model 的保持一致 (hold=1,2,3 in SAS). 可是
R2怎么这么小呢(< 1%), 几个 fixed effects 都非常 significant.
谢了先. |
|
c***z 发帖数: 6348 | 43 Weird, I found the Chinese version harder to understand...
If you are talking about panel data, here is a link
http://www.princeton.edu/~otorres/Panel101.pdf
In short:
The fixed-effects model controls for all time-invariant
differences between the individuals, so the estimated
coefficients of the fixed-effects models cannot be biased
because of omitted time-invariant characteristics…[like culture,
religion, gender, race, etc]
One side effect of the features of fixed-effects models is that
they c... 阅读全帖 |
|
e***7 发帖数: 862 | 44 工作中遇到的一个问题: 银行客户issue loan 的后续follow up,customer service会
打电话access criteria,大部分不会有改变,但有时会cancel(不符合标准) 或者是
increase rate,我们需要做一个model 把这些会改变的挑出来。
我用GLM可以model 'increased rate', logistic 可以model cancel, 这2个model
最后都需要score, 第一个model可以算出predicted rate increase,第二个logistic就
只有probability,如果要combine 这2个model,请问一般是怎么处理?
我想的是weight它们,但具体怎么给cancel这种情况赋上一个'increased rate'数值? |
|
y**3 发帖数: 267 | 45 this is time -to-event data. But we want to perform time-sliced model
instead of survival model. for instance, to build a model at 1 year to
predict the event or not. The set up of the data should censor all the data
beyond 1 year and then to identify event and predictors within 1-year and
then train the model. Not using data beyond 11 year to identify event or not
for the 1-year model?
Anybody confirm? thanks |
|
t*********p 发帖数: 44 | 46 【 以下文字转载自 JobMarket 讨论区 】
发信人: thebodyshop (body...), 信区: JobMarket
标 题: [招人]Quantitative Modeler
发信站: BBS 未名空间站 (Tue Apr 19 18:04:40 2016, 美东)
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 ... 阅读全帖 |
|
a******n 发帖数: 11246 | 47 听上去有点像是银行里做的retention model。
就用logistic regression就行了。training data把每个账单上的所有信息都记录下来
,可以加上些别的有的没有,比如你说的什么登陆频率。(btw,登陆频率,你用time
since last login, time since 2nd last login, etc...这样像你提到的那个人就是
比如,200天,201天,203天,etc...实战中用time since last login应该就够了,最
多加个time since 2nd last login)
然后这些货,以及他们的aggregation, transformation,统统扔到model里。什么
univariate analysis, cluster analysis, stepwise logistic regression,统统用
上,就出来model啦。这里题外话已经,我非常反对把某个变量在时间1和时间2的差放
入model。原因是,这个差和dependent variable肯定不是单调的,所以说不定会对
model反而... 阅读全帖 |
|
s********1 发帖数: 235 | 48 【 以下文字转载自 Statistics 讨论区 】
发信人: someone111 (some), 信区: Statistics
标 题: 有什么模型能把linear regression model 和 time series model 合起来做prediction 吗?
发信站: BBS 未名空间站 (Wed Dec 9 17:00:04 2015, 美东)
有什么模型能把linear regression model 和 time series model 合起来做
prediction 吗?现在有一些数据,是一堆产品,他们有一些自身的属性的数据,如网
上的ratings 等,他们还有time series 的销售数据,每一种产品有month by month
的销售值,现在要用这些数据做predictive modeling, 想到的方法有对自身属性的数
据可以做linear regression, svm 这样的模型预测,对time series 的销售数据,可
以用 time series 的模型预测,有没有什么模型能把这两种模型结合起来,用一个模
型考虑两方面,进... 阅读全帖 |
|
s********1 发帖数: 235 | 49 【 以下文字转载自 Statistics 讨论区 】
发信人: someone111 (some), 信区: Statistics
标 题: 有什么模型能把linear regression model 和 time series model 合起来做prediction 吗?
发信站: BBS 未名空间站 (Wed Dec 9 17:00:04 2015, 美东)
有什么模型能把linear regression model 和 time series model 合起来做
prediction 吗?现在有一些数据,是一堆产品,他们有一些自身的属性的数据,如网
上的ratings 等,他们还有time series 的销售数据,每一种产品有month by month
的销售值,现在要用这些数据做predictive modeling, 想到的方法有对自身属性的数
据可以做linear regression, svm 这样的模型预测,对time series 的销售数据,可
以用 time series 的模型预测,有没有什么模型能把这两种模型结合起来,用一个模
型考虑两方面,进... 阅读全帖 |
|
w*******y 发帖数: 60932 | 50 Here's the deal. There was a Black Friday deal on Model LCT2010 for $199.00
at Home Depot. There was short supply, and my Home Depot was sold out
quickly. I believe this model was packaged especially for Home Depot, since
you really can't find the LCT2010 model anywhere else.
BUT, model LCT306W (which Home Depot also sells, along with everyone else)
is essentially the same thing, and I decided to price check it in store at
Home Depot. The shelf price was $279, but it rang up $199. I used a 10% o... 阅读全帖 |
|