r******g
发帖数: 4002 | 1 On May 30th, I made a prediction that shocked the readers of TownHall. I
predicted a Romney landslide- with Obama leading in every poll.
I did not make that prediction as a political columnist, or as the former
Libertarian Presidential contender, or the 2008 Libertarian Vice
Presidential nominee. I relied on my career as a Las Vegas oddsmaker. Long
before I got into politics, I started out as the Network Oddsmaker and NFL
Analyst for CNBC (then known as Financial News Network). I've made my livi... 阅读全帖 |
|
r******g
发帖数: 4002 | 2 On May 30th, I made a prediction that shocked the readers of TownHall. I
predicted a Romney landslide- with Obama leading in every poll.
I did not make that prediction as a political columnist, or as the former
Libertarian Presidential contender, or the 2008 Libertarian Vice
Presidential nominee. I relied on my career as a Las Vegas oddsmaker. Long
before I got into politics, I started out as the Network Oddsmaker and NFL
Analyst for CNBC (then known as Financial News Network). I've made my livi... 阅读全帖 |
|
B*****e
发帖数: 2413 | 3 I want to sponsor a SPY prediction game for one month.
The rules:
1. Monday~Thursday prediction:
In the first day , make your prediction for SPY closing
point for the second day. Whoever's prediction is in the range of +/-
0.30 and also most close to SPY closing point of second day than any
other predictions, whoever wins 20 weibi.
2. Friday prediction:
In Thursday of every week, make your prediction for SPY closing point
for
the end of week. Whoever's prediction is in the range of +/-0.30 and
a... 阅读全帖 |
|
Y*****s
发帖数: 342 | 4
quality of
point of view (breeze, msee, bison, etc)
Wrong prediction:
result,QB DEF RB WR TE
Correct preditcion:
K
much better DEF. this should be a big margin win. (fwiw, froger
hasn't even adjusted his lineups yet)
Wrong prediction:
result,QB WR RB TE K DEF
correct prediction:none
Wrong prediction:result,QB RB,TE
correct prediction:DEF,K
wrong prediction:result,QB WR,DE, TE,K,RB
correct prediction:none
wrong prediction:result,K,RB,WR
correct prediction: TE,DEF,QB
wrong predicti |
|
c***z
发帖数: 6348 | 5 Anyone interested?
http://www.kaggle.com/c/march-machine-learning-madness
Predict the 2014 NCAA Basketball Tournament
Each year, millions of people fill out a bracket to predict the outcome of
the Division 1 NCAA Basketball Tournament. Whether you base your predictions
on published rankings, tourney seeds, media pundits, betting gurus, offense
stats, defense stats, an upset strategy, or the hottest trending teams, it'
s clear there are as many ways to predict the results as there are results
to ... 阅读全帖 |
|
D*****t
发帖数: 558 | 6 Every yr, the famous Byron Wien writes about 10 surprises that he predict to
happen. Meanwhile, Doug Kass, a hedge fund manager who shorts stocks, does
the same thing. Their predictions do not limit to market, but cover
politics also. One is not to take their predictions seriously. But I do
enjoy reading them very much, whether they are right or wrong.
This yr, Wien got it so wrong. For example, he predicted Iran gains nuclear
capability; SP500 decline below 1300; Crude falls to $70 while gold r... 阅读全帖 |
|
d******8
发帖数: 1972 | 7 http://us-non-immigrants.blogspot.com/2011/02/prediction-for-eb
Here is the Prediction for EB2 Category cutoff date movement for Fiscal Year 2011. Basis for this prediction is simple calculations (see below) which is done based on available data i.e. visa statistics recently released for CY 2010, PERM data published by DOL for CY 2010, latest trend on Trackitt for EB2 cases and other published data by USCIS. Demand data for each dependent category is predicted and explained in calculations below... 阅读全帖 |
|
d******8
发帖数: 1972 | 8 New Predictions - Last Updated - 01 May 2011
http://us-non-immigrants.blogspot.com/2011/02/prediction-for-eb
China
Optimistic: 6/1/2007
Realistic: 3/8/2007
Worst : 1/1/2007
Here is the Prediction for EB2 Category cutoff date movement for Fiscal Year 2011. Basis for this prediction is simple calculations (see below) which is done based on available data i.e. visa statistics recently released for CY 2010, PERM data published by DOL for CY 2010, latest trend on Trackitt for EB2 c... 阅读全帖 |
|
y*b
发帖数: 3190 | 9 "I think we’re all waiting eagerly for 2016, as we expect a lot of new
consumer, prosumer and professional products from Canon.
Below is what we expect to see from Canon with some predicted specifications.
Canon EOS-1D X Mark II
This is a no-brainer and we expect this to be the first big product
announcement from Canon in 2016.
Predicted specifications: 20-24mp, ISO 409,000, 15fps, 4K Video, DIGIC 7,
More than 61AF Points, Dual CFast, GPS
Canon EOS 5D Mark IV
We expect to see a brand new version... 阅读全帖 |
|
d******8
发帖数: 1972 | 10 http://us-non-immigrants.blogspot.com/2011/05/demand-data-for-j
Demand data that will be used for June Visa Bulletin has been released.
Please see below the cumulative demand reduction from May 2011 to June 2011.
Some of these reductions were as predicted and others were less due to
hidden demand (usually newly added cases from CP offices and regular
transfer of cases from local offices to National Service Center). Negative (
-ve) number indicates increase in demand. Based on the currently relea... 阅读全帖 |
|
d******8
发帖数: 1972 | 11 July 2011 Visa Bulletin Predictions - EB Category
http://us-non-immigrants.blogspot.com/2011/05/july-2011-visa-bulletin-predictions-eb.html
Sunday, May 22, 2011 | AOS , Employment-Based , Green Card , I-485 ,
Immigration , July 2011 , Predictions , Visa Bulletin
Here is the prediction for July 2011 Visa Bulletin. We should see following
movements in each category as long as USCIS/DOS would use visa numbers for
each category as per statutory allocations and as long as EB2-India and
China will con... 阅读全帖 |
|
g*****n
发帖数: 57 | 12 DOS Predictions: FB2A to Move Forward, EB2 India to Retrogress
July 15, 2014
The American Immigration Lawyers Association (AILA) has announced that it
will now begin to issue a report each month intended to provide reliable
predictions and analysis regarding the movement of cutoff dates in the visa
bulletin. These reports will be based on discussions with Charles Oppenheim,
Chief of the Visa Control and Reporting Division, U.S. Department of State
(DOS), following the release of each month's vis... 阅读全帖 |
|
p******m
发帖数: 65 | 13 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！ |
|
O*****n
发帖数: 78 | 14 我觉得你首先得确定你的modeling approach - 你打算做causal-predictive model还
是explorative-predictive model？前者必须根据theory来确定predictors，然后收集
数据，进行验证，如果结果满意，那么根据model进行prediction。后者可以根据现有
data去找出哪些predictors和预测对象有correlation，然后在根据data调整model，直
到获得满意的结果。
前者的model很容易解释，因为都是有现成的theory做基础的。但是如果theory有问题
，比如缺少了某些考虑，或者时间地点的变化导致其适用性不佳，那么model的
prediction可能很不好。
后者的model可能会难以解释，因为是从data里derive出来的，但是一般可以获得比较
好的prediction。
顺便提一句，如果你打算用SEM的话，用partial least squares对prediction比较好。 |
|
l******n
发帖数: 9344 | 15 Goldman Sachs 2011 Forecast: Stocks, Gold, Oil Higher
Goldman Sachs is bullish on the U.S. economy for 2011, and forecasts U.S.
stocks will see their third straight year of gains.
The investment banking powerhouse sees the S&P 500 [.SPX Loading... ()
] gaining nearly 25 percent to a level of 1450 in the next 12 months,
fueled by strong corporate profits, easy monetary policies and an improving
U.S. economy.
Goldman [GS Loading... () ] sees stocks gaining as the U.S. economic
gro... 阅读全帖 |
|
H*V
发帖数: 2770 | 16 dream ba
Year 2011. Basis for this prediction is simple calculations (see below)
which is done based on available data i.e. visa statistics recently released
for CY 2010, PERM data published by DOL for CY 2010, latest trend on
Trackitt for EB2 cases and other published data by USCIS. Demand data for
each dependent category is predicted and explained in calculations below.
This data is further used to calculate spillover that would b: e available
for EB2 category. Dates are predicted based on vi... 阅读全帖 |
|
m*******e
发帖数: 1280 | 17 his name is Iben Browning, he predicted the SF earthquake 1 week before it
happened.
His prediction on the STL area was supposed to be on Dec 3, 1990.
"Browning's earthquake prediction just 6,346 days late".
http://www.dailyherald.com/story/?id=175885
http://www.showme.net/~fkeller/quake/comments.htm
it was actually predicted to happen here in STL in 2000.
The guy who made the prediction already died, actually. His name is Brown,
as I remembered.
It is limited in 30 years now. |
|
n**a
发帖数: 26 | 18 2014的新年参加了LOCAL RUNNING CLUB的New year resolution 8K prediction run。
这个prediction run是这样的比赛方法，不管你什么时候开始跑，谁在最接近11：00AM
的时候跨过终点谁就是冠军。当然，前提条件是所有参赛者都不许带任何计时设备，只
能凭自己的感觉跑。
我在感恩节的时候参加过一次这个组织的turkey trot 10K prediction run， 结果在
10：50跨过终点，差了整整10分钟。
这次我对自己的速度有了更好的认识，我在10：10AM出发了，维持着自己舒服的10分每
迈的速度。在后半程不断的被很多人超过，我自己心里也没底到底是什么时间了。跨过
终点的时候，也不知道是几点，所有的人都没有表，没有手机。就算自己有表，和比赛
的标准时间差个十几秒也是常事。
比赛结果到晚上才在网上公布出来，我居然得了第3名，成绩是10点59分58秒，离11点
整只差2秒。这可真是dumb luck。我估计在以后的prediction run里也不可能有更接近
的成绩了。当然在真正的比赛里，我这样的龟速是不敢奢望任何名... 阅读全帖 |
|
m******t
发帖数: 273 | 19 【 以下文字转载自 Statistics 讨论区 】
发信人: myregmit (myregmit), 信区: Statistics
标 题: data prediction by regression or better ways
发信站: BBS 未名空间站 (Fri Mar 7 17:24:34 2014, 美东)
I am working on data prediction.
Given data of a random variable X and Y, find out how to predict Y by X.
I know how to do it by linear regression, y = k x + b .
But, here, x is always non-negative and y is required to be non-negative.
Sometimes, b is not non-negative so that y < 0.
How to assure that b > 0 and also minimize the pre... 阅读全帖 |
|
m******t
发帖数: 273 | 20 【 以下文字转载自 Statistics 讨论区 】
发信人: myregmit (myregmit), 信区: Statistics
标 题: data prediction by regression or better ways
发信站: BBS 未名空间站 (Fri Mar 7 17:24:34 2014, 美东)
I am working on data prediction.
Given data of a random variable X and Y, find out how to predict Y by X.
I know how to do it by linear regression, y = k x + b .
But, here, x is always non-negative and y is required to be non-negative.
Sometimes, b is not non-negative so that y < 0.
How to assure that b > 0 and also minimize the pre... 阅读全帖 |
|
m******t
发帖数: 273 | 21 【 以下文字转载自 Statistics 讨论区 】
发信人: myregmit (myregmit), 信区: Statistics
标 题: data prediction by regression or better ways
发信站: BBS 未名空间站 (Fri Mar 7 17:24:34 2014, 美东)
I am working on data prediction.
Given data of a random variable X and Y, find out how to predict Y by X.
I know how to do it by linear regression, y = k x + b .
But, here, x is always non-negative and y is required to be non-negative.
Sometimes, b is not non-negative so that y < 0.
How to assure that b > 0 and also minimize the pre... 阅读全帖 |
|
m******t
发帖数: 273 | 22 【 以下文字转载自 Statistics 讨论区 】
发信人: myregmit (myregmit), 信区: Statistics
标 题: data prediction by regression or better ways
发信站: BBS 未名空间站 (Fri Mar 7 17:24:34 2014, 美东)
I am working on data prediction.
Given data of a random variable X and Y, find out how to predict Y by X.
I know how to do it by linear regression, y = k x + b .
But, here, x is always non-negative and y is required to be non-negative.
Sometimes, b is not non-negative so that y < 0.
How to assure that b > 0 and also minimize the pre... 阅读全帖 |
|
e*****u
发帖数: 67 | 23 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. |
|
t*******1
发帖数: 135 | 24 【 以下文字转载自 DataSciences 讨论区 】
发信人: travis321 (travis), 信区: DataSciences
标 题: 急，xgboost prediction的问题
发信站: BBS 未名空间站 (Thu Nov 19 02:35:33 2015, 美东)
用r的xgboost package做prediction，是不是prediction dataset和model dataset每
一列的顺序一定要是一致的？为什么我的prediction dataset如果某个feature移动了
位置，比如从第一列变成了第三列（但是feature name不变），prediction的结果就不
一样了？
请用过的人帮忙解答一下。万分感谢啊！ |
|
o**********e
发帖数: 18403 | 25 【 以下文字转载自 SanFrancisco 讨论区 】
发信人: onetiemyshoe (onetiemyshoe), 信区: SanFrancisco
标 题: PREDICTION: 只要TPP不过，每两天抓中国人
发信站: BBS 未名空间站 (Sat May 23 11:20:54 2015, 美东)
发信人: onetiemyshoe (onetiemyshoe), 信区: USANews
标 题: PREDICTION: 只要TPP不过，每两天抓中国人 (转载)
发信站: BBS 未名空间站 (Sat May 23 11:22:06 2015, 美东)
发信人: onetiemyshoe (onetiemyshoe), 信区: SanFrancisco
标 题: PREDICTION: 只要TPP不过，每两天抓中国人
发信站: BBS 未名空间站 (Sat May 23 11:20:54 2015, 美东)
MY BOLD PREDICTION：
老美FBI/NAVY得到OBAMA的指示是：“
只要TPP不过，每两天造点黑中国的事情，
比如抓抓中国间谍，去南海... 阅读全帖 |
|
o**********e
发帖数: 18403 | 26 这半年了。 还是这句话。 O8 只要国会不通过TPP，
会整天挑衅老中南海，和抓老中间谍的。对策依然如下。
【 以下文字转载自 SanFrancisco 讨论区 】
发信人: onetiemyshoe (onetiemyshoe), 信区: SanFrancisco
标 题: PREDICTION: 只要TPP不过，每两天抓中国人
发信站: BBS 未名空间站 (Sat May 23 11:20:54 2015, 美东)
发信人: onetiemyshoe (onetiemyshoe), 信区: USANews
标 题: PREDICTION: 只要TPP不过，每两天抓中国人 (转载)
发信站: BBS 未名空间站 (Sat May 23 11:22:06 2015, 美东)
发信人: onetiemyshoe (onetiemyshoe), 信区: SanFrancisco
标 题: PREDICTION: 只要TPP不过，每两天抓中国人
发信站: BBS 未名空间站 (Sat May 23 11:20:54 2015, 美东)
MY BOLD PREDICTION：
老美... 阅读全帖 |
|
l****z
发帖数: 29846 | 27 August 7, 2012 Posted by David Robertson
The government of North Carolina is being mocked by believers in
anthropogenic global warming because the state’s Democrat governor
permitted a controversial legislative bill to become law.
North Carolina House Bill 819 prohibits state officials from relying on
current computer models to predict what the sea level along the state’s
coastline will be in the future.
An ABC News report about the controversial law states the following:
Republican State ... 阅读全帖 |
|
m****s
发帖数: 18160 | 28 【 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 |
|
s****2
发帖数: 4569 | 29 From News.
The 'Psychic Octopus' paul has predicted a German World Cup Win this Sunday.
This isn't just any octopus, it's a 2-year-old octopus prophet that was
born in England, but raised in Germany. Does he really have the ability to
predict Soccer game outcomes?
We are already suspicious of this 8 legged sea monster. How fitting that a '
Psychic Octopus' in Germany would predict a German victory of England on
Sunday.
The Psychic Octopus's name is Paul. Paul also predicted a German victory
over |
|
o**********e
发帖数: 18403 | 30 【 以下文字转载自 SanFrancisco 讨论区 】
发信人: onetiemyshoe (onetiemyshoe), 信区: SanFrancisco
标 题: PREDICTION: 只要TPP不过，每两天抓中国人
发信站: BBS 未名空间站 (Sat May 23 11:20:54 2015, 美东)
发信人: onetiemyshoe (onetiemyshoe), 信区: USANews
标 题: PREDICTION: 只要TPP不过，每两天抓中国人 (转载)
发信站: BBS 未名空间站 (Sat May 23 11:22:06 2015, 美东)
发信人: onetiemyshoe (onetiemyshoe), 信区: SanFrancisco
标 题: PREDICTION: 只要TPP不过，每两天抓中国人
发信站: BBS 未名空间站 (Sat May 23 11:20:54 2015, 美东)
MY BOLD PREDICTION：
老美FBI/NAVY得到OBAMA的指示是：“
只要TPP不过，每两天造点黑中国的事情，
比如抓抓中国间谍，去南海... 阅读全帖 |
|
d******k
发帖数: 4295 | 31 NBA赢了，把本来一边倒的比赛拖到第6场。
We’re sure that this year’s Golden State Warriors-Cleveland Cavaliers NBA
Finals won’t beat the average of the 1998 Utah Jazz-Chicago Bulls finals.
That series, Michael Jordan’s last in the NBA Finals, averaged 29 million
viewers. But will at beat everything since then?
We think so. The only series standing in its way is 2001’s Lakers-76ers
series that averaged 18.996 million viewers on NBC.
By winning Thursday’s game, Golden State deadlocked the series at two games
apiece... 阅读全帖 |
|
H****w
发帖数: 1632 | 32 One 包子 for your Super Bowl prediction.
One prediction per person.
First 20 people with prediction only.
>Which team will win the Super Bowl on 02-02-2014?
Denver or Seattle?
恭喜发财 and 谢谢 ~~~ |
|
m****s
发帖数: 675 | 33 denver
One 包子 for your Super Bowl prediction.
One prediction per person.
First 20 people with prediction only.
>Which team will win the Super Bowl on 02-02-2014?
Denver or Seattle?
恭喜发财 and 谢谢 ~~~ |
|
R******d
发帖数: 1436 | 34 【 以下文字转载自 Statistics 讨论区 】
发信人: Rainbird (落汤鸟), 信区: Statistics
标 题: 想问一个关于评价prediction performance的问题
发信站: BBS 未名空间站 (Tue Oct 26 13:06:46 2010, 美东)
现在想做一个predictor，用来预测人群中的某种不常见的疾病，发病率不到1%。
training data是
非常imbalanced的，positive data points很少，绝大部分都是negative data points
。我
没有直接用这样的training data，而是人为地构建了balanced data。简单的说，就是
保持
positive data points不变，随机选同样sample size的negative data points。重复
训练
若干次，最后的训练结果是这么多次结果的汇总。
因为发病率确实很低，所以我取的specificity很高，比如99.9%。相应来说
sensitivity就很低
了，2%不到。换算成Positive Predic... 阅读全帖 |
|
v**n
发帖数: 130 | 35 unlikely.
predictive modeling deals with what the theoretical answer should be.
Actuaries have to use their business judgement and figure out what the best
solution for the real life business problem is.
I don't see actuarial capabilties be replaced by modeling completely... we
are a long way away from using statistical models to predict the real world
accurately....
however, predictive modeling is a skill that newer actuaries should master.
it will be very hard to have a secure career going for... 阅读全帖 |
|
s*********e
发帖数: 1051 | 36 【 以下文字转载自 Statistics 讨论区 】
发信人: statcompute (statcompute), 信区: Statistics
标 题: Predictive Analytics Developer
发信站: BBS 未名空间站 (Tue Sep 3 11:04:54 2013, 美东)
the client is based in Chicago they would be open to having someone work
remotely as long as they are open to travel.
Sonia Macias
Assistant Cindy South
SC Riverside Inc.
712-325-6884
712-325-6691 (F)
www.scriverside.com
http://www.linkedin.com/in/soniamacias
Summary
The Predictive Analytics Developer is an expert in the statistical methods
... 阅读全帖 |
|
s*****n
发帖数: 2174 | 37 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. |
|
s*****n
发帖数: 2174 | 38 我不觉得有什么放之四海皆准的程序可以使你降低prediction error.
你唯一能做的, 就是尝试不同的variable selection, 尝试不同的transforation.
如果你的reponse是近似normal的, 尽量把你所有的predictor都往normal上面
transform. 如果response非常skew, 你首先要把response变得近似normal了, 至少也
要比较symmetric了.
还有一点就是, 我不知道你是如何选model和评价prediction的. 如果你没有用cross-
validation的话, 最好用这个标准. 或者是用AIC做标准也一样, 理论上, AIC 是试图
minimize prediction error 的. 仅仅看 R^2 这些来试图找到predictive model 肯定
是不行的. |
|
s*r
发帖数: 2757 | 39 for proportional data, it could be predictive value in stead of predictive
probability. google positive predictive value negative predictive value.
to raise power without increasing sample size, you have to raise type I
error bar |
|
m******t
发帖数: 273 | 40 I am working on data prediction.
Given data of a random variable X and Y, find out how to predict Y by X.
I know how to do it by linear regression, y = k x + b .
But, here, x is always non-negative and y is required to be non-negative.
Sometimes, b is not non-negative so that y < 0.
How to assure that b > 0 and also minimize the prediction error ?
Are there other better ways (not regression) to do the prediction ?
Any help would be appreciated. |
|
m******t
发帖数: 273 | 41 I am working on data prediction.
Given data of a random variable X and Y, find out how to predict Y by X.
I know how to do it by linear regression, y = k x + b .
But, here, x is always non-negative and y is required to be non-negative.
Sometimes, b is not non-negative so that y < 0.
How to assure that b > 0 and also minimize the prediction error ?
Are there other better ways (not regression) to do the prediction ?
Any help would be appreciated. |
|
h***i
发帖数: 3844 | 42 你不是自己解答了么，predictive model，
如果只在乎model 的prediction的效果，那么放不放就看你prediction的结果好还是不
好。
如果还在乎model要简洁，干净，那就牺牲一点prediction power，弄个简单的没
interaction 的就好了。 |
|
w*********y
发帖数: 7895 | 43 SEM挺好用的， 一般要大于200人的数据（我不知道MAX NUMBER), 然后TEST一下SEM的
ASSUMPTIONS，无非就是homogeneity之类的。
然后你想好自己的prediction, what factor predicts future income ,what are the
moderators, what are the mediators,
然后看MODEL FIT..如果是SIGNIFICANT，说明PREDICTION成立。。
比较简单的方法是，从最SIMPLE的PREDICTION开始，比较各种因素对
结果的预测。。。就这样。。。、
HLM就比较复杂，我自己也还在学习。。。 |
|
m******a
发帖数: 77 | 44 大致作法应该是这样的
1. 你有一列数据点(y0, x0), 这里x应理解为N维矢量
你的模型是 y = f(x, c), c 是一列模型参数
在你建模型时, 你是要定出这一列参数 c, 同时你还应该得到对应的 Covariance
Matrix
通常的软件都提供这些信息
2. 有了Covariance Matrix后, 你可以生成很多列参数 c,
下面实际是作一些 Monte Carlo 模拟,
对应每一个数据点, 你都有很多 prediction - 因为每列c 给一个prediction
因此, 如果你有 D 个数据点的话,
你得到 D by D 的一个Covariance Matrix M -别和前面的Covariance Matrix混淆
3. 现在你可以算出来 Chi-Square
SUM( (y0 - y) M^-1 (y0 - y) )
同时, D 就是你的自由度数
有了这两个, 查表就知道你的 confidence level, 就是你的模型有多可靠
4. 关于你说的 confidence interval... 阅读全帖 |
|
t*******1
发帖数: 135 | 45 用r的xgboost package做prediction，是不是prediction dataset和model dataset每
一列的顺序一定要是一致的？为什么我的prediction dataset如果某个feature移动了
位置，比如从第一列变成了第三列（但是feature name不变），prediction的结果就不
一样了？
请用过的人帮忙解答一下。万分感谢啊！ |
|
p****r
发帖数: 104 | 46 【 以下文字转载自 Biology 讨论区 】
发信人: poster (第二性征), 信区: Biology
标 题: New exciting research! Prediction of quantitative phenotyp
关键字: prediction,phenotype,genetic network
发信站: BBS 未名空间站 (Mon Sep 13 21:03:04 2010, 美东)
A group of UCSD researchers have just provided an approach to make
prediction
of phenotypes of an organism from genetic networks. The method does not rely
on
any training data and can be based on literature curated or computationally
predicted networks. The method seems to be very easy to i |
|
a***o
发帖数: 3734 | 47 西医用药做手术的结果基本可以预知，少数人的结果可能有偏差，但是统计意义上西医
有predictability。中医压根没有predictability，你听到的都是个例，某某老中医治
好了某某人的某某病之类的。
人命关天，只有死马当活马医的人才去用没有predictability的东西。 |
|
m******n
发帖数: 6327 | 48 http://blogs.wsj.com/economics/2008/07/02/want-to-predict-olympic-champs-look-at-gdp/
Want to Predict Olympic Champs? Look at GDP
Back in early 2000, Daniel Johnson, a professor at Colorado College, found
himself with extra funding leftover from a separate project and with the
help of an undergrad decided to use the resources to see whether economic
variables could predict the medal winnings of each country in the upcoming
summer Olympics.
medals_cs_20080702102207.jpg
Click image for full list o |
|
l***g
发帖数: 1035 | 49 what's the point of these 'predictions'? the sample is too small to be repre
sentative. and yet there is always a 50-50 chance to get the prediction righ
t.
this is like trying to predict which side of the tossed coin up by analysing
the difference in weight on each side.
( |
|
S******8
发帖数: 24594 | 50 http://www.conradmblack.com/1221/november-8th-predictions
Before outlining my predictions for November 8th, let me first offer this
disclaimer: I am no statistician, do not consult computer algorithms, and
have no experience as a pollster. I am happy to leave all of that to
professionals like Nate Silver. However, my experience as an author of
American history, biographer of two American presidents (who between them,
ran ten times for national office from 1920 to 1972), and as someone who has
cl... 阅读全帖 |
|
