a*********d
发帖数: 2763 | 1 I actually wrote a letter to my previous residency program director to share
my scores and experience, he distributed this to all the IM residents. here
is the copy of my scores, and i posted my board experience in the past
already.
my overall performance is 617, (standardized passing score is 370).Decile 9
(575-617)Decile 10 (618-800) i only missed one point to be at Decile 10,
darn it!
i listed my score by specialty here:
Decile 10
* hematology 100% 12 questions
Decile 9
* endo 94% , i know |
|
e****0
发帖数: 378 | 2 take home assignment里的一道。由于我assigment做挂了，所以没机会知道反馈了。
想知道有什么好的写法。
table:
row_id - A unique id for the row
customer_id — A unique id for the loan borrower
loan_amount — The amount of the loan requested by the borrower
addr_state — Borrower's state of residence
问题：
Now let's rank the states by their top decile of loans to see where
customers borrow the most. Find the states where the cutoff for the top 10%
of loans (by loan amount requested) is the highest. If two states have the
same cut-off for t... 阅读全帖 |
|
a***a
发帖数: 40617 | 3 低碳水高蛋白饮食导致死亡率上升，公主看看，这篇文章居然都能发出来，harvard
水平坑爹啊。欢迎来找茬。lol
http://www.ncbi.nlm.nih.gov/pubmed/17391111
Display Settings:
Abstract
Send to
J Intern Med. 2007 Apr;261(4):366-74.
Low carbohydrate-high protein diet and mortality in a cohort of Swedish
women.
Lagiou P, Sandin S, Weiderpass E, Lagiou A, Mucci L, Trichopoulos D, Adami
HO.
Source
Department of Hygiene and Epidemiology, University of Athens Medical School,
Goudi, Athens, Greece. p*****[email protected]
Abstract
OBJECTIVE:
The long-term he... 阅读全帖 |
|
S******y
发帖数: 1123 | 4 Thanks for your elegant R code. Songkun!
To : DaShagen,
What I am doing is to cut data into deciles based on sorted variable of
interest (say score); and once I get deciles, I can compute means of
dependent variable across deciles - to see the lift. |
|
S******y
发帖数: 1123 | 5 I have a sorted data in R. I am trying to cut tiles/deciles based on
cumulative weights.
My code is below -- but how can I get rid of this loop (which might not be a
good coding practice here) ?
Thanks.
#------------------------------------------
#dat - R data frame (already sorted by variable of interest)
#n - number of obs
#tile - a vector of length n showing which tile the obs belongs to
#n_tile - number of tiles (i.e. deciles if it is 10)
#--------------------------------------
#R code -- |
|
h***x
发帖数: 586 | 6 According to my understanding to your question, you are saying that in the
top 2 deciles and bottom decile, there are lots of targets (you name it case
) while in the middle, fewer targets.
The distribution of the targets is based on the scores. The reason why lots
of targets are in the bottom is because they have low scores. So you need to
check the variable values for those records.
As you said, you have a lots of zero, that should explain the reason.
reason? |
|
s****u
发帖数: 1200 | 7 还是分十个decile,算每个decile predicted avg, actual avg
★ 发自iPhone App: ChineseWeb 7.8 |
|
t*******a
发帖数: 4055 | 8 法国五百万上层中产成税改大输家
发布时间： 2017-10-15 23:41:10 来源：欧洲时报 作者：周文仪编译 浏
览次数：51 评论：0
【欧洲时报周文仪编译】据《费加罗报》10月15日报道，法国500万上层中产阶级家庭
是政府税务改革的大输家，其生活水平将下降。
部长标榜“全民收益”
经济部长勒梅尔三周前公布政府2018年预算案时声称这套预算案“全民受益”，但甚至
经济部的单位都承认：新的财政法案“不是毫无例外人人受益”。明年一些富裕家庭的
生活水平将下降，据财政法案中的一份附属文件，生活水平下降明确集中在法国人口中
收入属于第九个十分位的家庭。
第二档的“富人”是输家
统计学用“十分位”（DECILE）的概念把法国人口分成十个档次，每个档次为10%，包
括280万个家庭。最富的档次是第十个，包括法国最富的10%的富豪，排名第二的是收入
较丰的第九个十分位的家庭。据国家统计及经济研究所（INSEE）的统计，所有的家庭
类型混合，这些纳税户一年不包括直接税的净收入介于49350和63210欧元之间，单身一
人年收入介于27000和34000欧元之间，育有两... 阅读全帖 |
|
发帖数: 1 | 9 最近，在美国的一些中文网站上，兴起了一阵“老留”针对“小留”的讨伐。老留，严
格意义上说，是指2000年大规模自费留学开始之前，到达美国，落地生根，如今事业有
成，子女正在陆续进入大学和婚恋年龄的一部分人。小留，特指在2010年前后，自费来
美国读高中和大学本科的一部分年轻人。老留和小留可以说不是一代人，甚至可以说，
小留和老留的孩子，即所谓ABC, “华二代” 是同一代人。
老留的心态，有几分倚老卖老，也有几分自亮伤疤的伪豪迈，因为他们最喜欢的就是说
自己当初如何怀揣几十美元到了美国，一下飞机就进餐馆打工，或者第二天就进了实验
室干到晚上十点，以此反衬小留如何依赖父母，靠父母养活，花天酒地，不爱读书，习
惯不良等等，同时吹嘘自己的孩子如何成功“爬藤”——进入美国的常春藤盟校，如何
学钢琴，如何获奖。这种沾沾自喜的心态，已经成为现在五十多岁左右的在美“老留”
中普遍流行的症状。
其实，“老留” 这一代人仔细分析起来，未必是完全经得起推敲的，虽然他们已经极
为习惯在一个封闭的小圈子里以“精英”自命。首先看一些数据：“90年代以来，尤其
是2000年后，随着留学不再受限以及人们收入增加，留... 阅读全帖 |
|
l****z
发帖数: 29846 | 10 By Michael Barry
I keep thinking about Joe the Plumber. I went back and read the full
transcript of his dialogue with then-candidate Obama, in October of 2008,
right when all the awful things had begun to happen to the economy. It is
really an amazing exchange of views -- most amazing because no other
interviewer or media question-shouter ever got as much out of Obama about
his economic vision as Joe did. When Tom Brokaw later complained on Charlie
Rose that "We don't really know anything abo... 阅读全帖 |
|
l****z
发帖数: 29846 | 11 On Labor Day, 1980, Jimmy Carter’s presidential re-election campaign was
ahead by 4 points. On November 4, 1980, the incumbent lost by 10 points. As
I learned firsthand back then, a lot can happen in two months.
Today, as the RealClearPolitics average shows Mitt Romney down by less than
a point in his White House bid, the challenger needs “the fierce urgency of
now,” to borrow a phrase often used by Barack Obama in 2008--although he
hasn’t used it much since. Ronald Reagan had that sense of ur... 阅读全帖 |
|
m********a
发帖数: 1041 | 12 Unequal protection
by From the print edition: Briefing
April 25, 2013
WILLIAM POWERS is the president of the University of Texas at Austin (UT-
Austin). Lino Graglia holds an endowed chair at its law school. Both have
kindly demeanours, impressive records and that crucial perk of academic
success, offices with great views: Mr Powers looks out over the heart of the
university’s campus, Mr Graglia at its football stadium.
They also hold strong and opposing opinions on whether admissions to their
... 阅读全帖 |
|
b********n
发帖数: 38600 | 13 Sanders and Trump: How the Political and Media Establishment Got 2016 So
Wrong
http://historynewsnetwork.org/article/162019
Perhaps the biggest story coming out of campaign 2016 is not the rise of
Bernie Sanders and Donald Trump, but the fact that the media and political
establishment never saw it coming. And the fact that they never saw it
coming perfectly explains the rise of Sanders and Trump.
And here's why: the economic and social wreckage wrought by the Great
Recession of 2007-2009, which ... 阅读全帖 |
|
j****h
发帖数: 219 | 14 OAKBX 更好，
Trailing 3-Yr through 04-30-10 | *Trailing 5-Yr through 04-30-10
Standard Deviation 11.53
Mean 4.32
Sharpe Ratio 0.24
Bear Market Decile Rank* 1 |
|
f**********n
发帖数: 29853 | 15 论文找到了
Abstract
Texas engaged in a large-scale policy experiment when it instituted the Top
10% Plan. This policy guarantees automatic admission to their state
university of choice for all high school seniors who graduate in the top
decile of their high school class. We find evidence that households reacted
strategically to this policy by moving to neighborhoods with lower-
performing schools, increasing property values in those areas. The effect is
strongest among schools that were very low-per... 阅读全帖 |
|
b*******e
发帖数: 24532 | 16 在申请NIW里面属于bottom 10~30 decile 吧，但是不代表没可能啊。也看看别的同学
评价。 |
|
z****g
发帖数: 3509 | 17 学历和收入的统计数据不好找，那就来看看英语。
英语拿满分没那么容易（包括对印度人来说），这也是我从办澳洲和加拿大永居中知道
的。
这个法案初稿里面，英语要拿满分需要在top tenth decile （也就是90%的人以上），
这个标准换算成托福分，如果读听说写分开算，分别需要
29 29 27 28
如果只看总分，需要108
注意这不是当初那个600多分里只有听和读的托福了。从澳洲和加拿大的制度看，最终
读听说写分开算的可能性很大。读听能拿29的我相信这里很多，说写要拿27和28，这里
的人有这个自信的有多少？特别是信奉“语言只是个工具”的那些老中，说写通常不达
标。
其实对印度人也是一样，你别看他们说得溜，他们的口音在工作中别人习惯了可能不成
为他们的障碍，但真去考试是要扣分的。另外如果你看过印度人写的东西就知道，他们
冠词使用经常出错，而且还会写出一些印度英语才有的用法，这都是要扣分的。从澳洲
和加拿大的情况看印度人的英语并不比中国人明显占优。
所以回到最初的总结，那还是分高的没那么多。
提高英语要求，来强迫过来的老中提高英语，改善老中形象，也是对大家都有好处的事
。 |
|
b*******r
发帖数: 6655 | 18 GPA 4.0的全年级不会只有一个吧？ 这里一般没有individual rank, ranking is
reported by decile only, i.e. 1= top 10%. |
|
j****y
发帖数: 1714 | 19 对长寿可能不利。
长肌肉要不少蛋白质，一般担心是吃多了蛋白质可能增加癌症概率。
但是最新研究似乎没有发现low carb, high protein diet 和癌症的关联。
http://www.nutritionj.com/content/12/1/58
Low-carbohydrate, high-protein diet score and risk of incident cancer; a
prospective cohort study, 2013
Abstract
Background
Although carbohydrate reduction of varying degrees is a popular and
controversial dietary trend, potential long-term effects for health, and
cancer in specific, are largely unknown.
Methods
We studied a previously established low-carbohydrate, hi... 阅读全帖 |
|
P****D
发帖数: 11146 | 20 Decreasing carbohydrate or increasing protein intake by one decile were
associated with increase in total mortality by 6% (95% CI: 0-12%) and 2% (95
% CI: -1 to 5%), respectively.
既然你已经都看完了，能否说一下他们这个decrease/increase是多大幅度？你一天吃
他两斤纯蛋白质，那绝对提高死亡率…… |
|
m******r
发帖数: 1033 | 21 有道理。 你的讲解让我想起不久前看过一段视频。 是采访著名cart发明人breimann还
是frimann讲他当年怎么发明决策树算法的故事。 他当年是stanford搞物理的， 原本
用k_nearest_neighborhood 方法解决高能粒子classification的问题, 后来有天早上
在床上想啊想啊( 我也爱赖在床上想一些自己的土算法， 很有效)忽然就从knn联想到
了决策树算法。 后来创办了CART软件，不久前被minitab收购。
现在我知道应该怎么向客户展示了： 应该类似和散点图(scatterplot)差不多的图，重
点展示一些cluster。
还有一个疑问。 xgb输出是score, 0~1， 我然后根据分数排序, 看看每个十分位(
decile)，购买率是多少， 结果竟然是100%, 99% 90%, 80%, 30%， 12%， 10%， 0%,
0%, 0% （就是做lift curve 类似方法）
我再仔细研究score, 发现这些分数跳跃性很大， 有很多分数相同(up to 10th
decimal place).
怎么看怎么别扭。 我的理解是： 这... 阅读全帖 |
|
M*******r
发帖数: 165 | 22 1. For the US stock universe what is a good information coefficient for a
quant model?
2. What is an acceptable IC?
3. How would a size of the stock universe affect threshold values for
questions 1-2?
4. A stock selection factor is tested on a monthly data for the in-sample
period between June of 2005 through July of 2007 showing a good information
coefficient and decent decile spreads. What can we conclude about the
usefulness of this factor on a standalone basis going forward? List your
concer... 阅读全帖 |
|
S******y
发帖数: 1123 | 23 why the results from the following code leading to unequal decile sizes?
d1<- cut(1:10000,10,1:10)
table(d1)
d1
1 2 3 4 5 6 7 8 9 10
992 1002 1002 1002 1002 1002 1002 1002 1002 992 |
|
s*********e
发帖数: 1051 | 24 statistician in different industry looks at different measures.
for risk modeling, the standard measures include but are not limited to KS
statistics, ROC, gini co-efficient, and divergence. for credit scoring, PDO
is also a measure for predictiveness.
for marketing, it is different story. they look at the lift at the top
decile. |
|
h***i
发帖数: 115 | 25 最近学习做Response Model, Logistic Regression做完, 得到 score 后test model好
不好, 好像一般都用Cum Lift Chart比较no model and with model. 有没有大虾能给
个做10 deciles of Cumulative Lift Chart SAS code? 多谢多谢 |
|
s****u
发帖数: 1200 | 26 怎么判断prediction bad? 看Individual level? 看Decile 的情况？这里可以人为决
定怎么看，模型就有可能好，可能坏
★ 发自iPhone App: ChineseWeb 7.8 |
|
j**********3
发帖数: 305 | 27 baseline will be developmental sample. Decile the dev sample by score. Then
apply the same scoreband on your recent thru the door sample. Base on cum
volume for dev sample and recent sample. You can calculate KS and PSI. Then
you apply the same method on each individual characteristic in ur Acq model
.
Accept
baseline |
|
