d*******1
发帖数: 293 | 1 if normal distribution with mean =1 and std=2, then for a value =3, how can
I test it is significantly larger than 2 on 5%? |
|
l*******r
发帖数: 407 | 2 From the information your provided, there should be no statistical
difference between 3 and 2, because they are only 2 samples from that normal
distribution. The only difference is that two numbers have different
probabilities |
|
p******r
发帖数: 1279 | 3 我发现我的Y变量本身就严重right skewed，用了log变换了，做出来normality test还
是fail。 用box－cox会好点吗？
weighted LS是不是就是robust regression啊？ sas里面用什么proc呢？ 谢谢！！ |
|
l***a
发帖数: 12410 | 4 any normality tests are better than jarque-bera? and why? |
|
p**5
发帖数: 2544 | 5 let's say, A and B are independent and normal distribution.
If means and standard deviations of A and B are known, how to estimate the
variance of A/B? |
|
e****t
发帖数: 766 | 6 If X and Y are Bivariate normal distribution
P(X>a | Y>b) , does it have a closed form ?
I guess no. |
|
o******e
发帖数: 1001 | 7 X=(x_1,x_2,...,x_n) 是multvariate normal distribution,他们之间的相关系数是确
定的，现在的问题是如果观测到其中(n-1)个x的值，怎么预测第n个的分布，我们可以
用条件概率来计算，比如：
P(x_1|x_2,...,x_n)，但是这样的条件概率只能用数值计算吧，不然很难解。是不是还
有另外的方法？ |
|
o******e
发帖数: 1001 | 8 可能我讲的不清楚。比如trivariate normal distribution X={x_1,x_2,x_3},如何求：
p(x_1>K_1|x_2>K_2,x_2>K_3)=p(x_1>K_1,x_2>K_2,x_2>K_3)/p(x_2>K_2,x_2>K_3)
但是求p(x_1>K_1，x_2>K_2,x_2>K_3)时，只能用数值积分了？ |
|
p******r
发帖数: 1279 | 9 use box-cox method. A well chosen lamda will give you more "normality".
SAS has an option to apply box-cox method. |
|
|
|
o******e
发帖数: 1001 | 12 比如P(X>x,Y>y,Z>z),X,Y,Z是trivariate normal distribution,能有近似解析解吗？ |
|
d********p
发帖数: 31 | 13 有一组遗传学数据， x-axis 是染色体位置， y-axis 是相应的数值， 理论上y值是恒
定的（＝0）， 除非某染色体位置有某种遗传学功能，这时会有峰值出现。我要找的就
是这些峰。现在我的数据在图上显示一些local trends。我想去除这些trends，使得总
体上均值是straight line。我的问题是：
1. 是用loess还是local mean 做 normalization， 用什么R package？
2. 如何计算每个峰显著性的p-value？t-test, resampling
多谢回复。 |
|
n*****n
发帖数: 3123 | 14 It is called singular normal when the covariance matrix is not full ranked.
It is not a good counter-example.
im |
|
T*******I
发帖数: 5138 | 15 我直觉地认为此类命题和证明是出自经典的数学思维，在纯粹的统计学里可能毫无意义
，因为我们常常遇到这样的情形：X和Y以及Y和Z之间存在显著的线性关系，而X和Z之间
则未必一定存在显著的线性关系。那种认为X和Z之间因此而一定存在显著线性关系因而
试图证明这个命题从而避免任何抽样估计的必要性的人恐怕还没有走进统计学的殿堂。
因此，如果X和Z之间没有显著的线性关系，讨论甚至试图证明X和Z之间是否存在
bivariate normal distribution就是毫无意义的事情。
关于随机变量及其相互间的关系，只有样本估计一条路可走。 |
|
s*********y
发帖数: 34 | 16 一般a Normality Test 在SAS有3种方法: Anderson-darling 是其中一钟,需要样本量
在100以上. 如果我现在有一个比较大的population (1,000,000), 为了检测
population是否服从正太分布, 应该抽取多少样本合适?
因为population 很大,需要重复几次抽样来确定结论吗? 如果重复几次, 有几次的样本
否定正太分布, 另外几次的不否定正太分布, 如何结论? |
|
n**********e
发帖数: 18 | 17 dv不normal，有一个numeric的iv和多个categorical的iv
如果用的话，怎么判断结果有多少可信度？
如果不用这个，可以用哪个proc或test来替代？
如果还要做model selection的话，要用哪个？
谢谢！ |
|
n**********e
发帖数: 18 | 18 是的，residual不normal。
如果不用transform呢？是不是只好用Kruskal Wallis test了？那后面要model
selection怎么办呢？
transform |
|
h***i
发帖数: 3844 | 19 Kruskal Wallis test 也不能让residual normal吧
不transform就用generalized linear model |
|
n**********e
发帖数: 18 | 20 只是想知道这个dv和各iv的关系，最后还想做个model selection来筛选一下，目前就
是用的proc glm，但是又有normality的问题
如果用Kruskal Wallis test也可以分析出dv和各categorical iv的关系，但是后面的
model selection就不知道怎么做了。。。。。。 |
|
f********k
发帖数: 111 | 21 Just use Anderson–Darling test under the case when both mean and var are
known. It's more powerful than K-S for normality test.
data |
|
m******t
发帖数: 273 | 22 I am working in python 3.2 and matplotlib.
I need to plot y and x.
The numbers of x-axis are between 0 and 1000000.
In my plot, the numbers of x-axis are long and overlap each other.
I want to re-scale and normalize them between 0 and 10 and also show the
unit (such as 10^6) around the x-axis only once.
Eexample,
0, 1, 2, 3, ..., 10 (10x6)
How can I do that?
Any help would be appreciated. |
|
E***n
发帖数: 308 | 23 请问不是normal distribution 的话，是不是不可以用 I-MR chart? |
|
m***c
发帖数: 118 | 24 一个数据有三个变量pct1,pct2,pct3，均是工人工作完成率0 - 1，obs大概有1500个,
pct2和pct3分别有10多个missing values。已经检测过这三个变量均不是normal。
data a;
input pct1 pct2 pct3;
cards;
0.2345 0.2657 0.3410
0.8009 0.7011 0.6945
1.0000 0.5699 0.8940
0.7109 . 0.6945
0.5470 0.9999 1.0000
0.8901 0.5557 .
0.4522 0.9672 0.6012
。。。。。
；
run;
问题：
1. 一把情况下，如何比较pct1,pct2,pct3是不是significantly same or different?
2.如果后2个变量oct2,pct3的工人是通过第一个pct1的工人PSM（propensity score
matching）找出的结果， 比如，用工人1的年龄技术经验，，来找到匹配的工人2（工
人1和工人2具有很高的相似性），工人2的完成率就是pct2，同样以工人1再... 阅读全帖 |
|
w********m
发帖数: 1137 | 25 regression 的assumptions里面没有normality。 这个应该是基本常识吧。 |
|
l**********8
发帖数: 305 | 26 在实际工作中，同事确实是只plot看一眼，如果不是skewness太厉害，就算正太了。。
。而且只是看histogram 和density。 在建立linear Regression model之前，不是要
test一下 residual的normality 吗（或是看看qqplot）？ |
|
t*****a
发帖数: 459 | 27 那些normality test往往结论不一致，而且每一种都不是业界公认的标准。有时候可以
用某个test看一眼，如果p<0.01就比较警惕，如果p=0.05左右但是看着数据分布和正太
差不远也还是可以当做正太的。有一些东西如果不是正太就要按non-parametric的方法
来处理，但是往往数据和正太差不远的时候non-parametric和parametric的结果也差不
多，所以不需要太纠结。 |
|
y*****8
发帖数: 39 | 28 linear regression is actually least square which does not need normality. |
|
y*****8
发帖数: 39 | 29 linear regression is actually least square which does not need normality. |
|
y*****8
发帖数: 39 | 30 linear regression is actually least square which does not need normality. |
|
y*****8
发帖数: 39 | 31 linear regression is actually least square which doesn't need normality. |
|
i**********n
发帖数: 217 | 32 what is your purpose?
You can either normalized variables first,
or use cos like distance. |
|
l*****w
发帖数: 81 | 33 我的purpose 是要normalize euclidean 的。
主要有一些数据，要做一个打分系统，根据一些商品的attributes确定一组商品的
scoring.想把所有的score 放在一个interval 里。这样有可比性。
想了一下用指数函数，可以达到这个目的。
exp(-x), 0
就可以把一组0到无穷的数，放到0-1区间里。
当然有更好方法吧. |
|
m*****s
发帖数: 2227 | 34 would you add Northeast normal university in 校友联谊? Thank you |
|
t**g
发帖数: 522 | 35 What should be normal TSH reference level? 0.3 to 3 mIU/L or 0.3 to 5mIU/L |
|
|
w*******y
发帖数: 60932 | 37 This isn't as good as the TransLink deal I posted a while ago where you got
a preloaded card with $5 on it, but for a limited time there is no charge to
get a Clipper card. "Normally $5, the fee for a card is being waived
starting June 16. This offer is available for a limited time and may be
subject to change without notice."
As you may or may not know Clipper has replaced Fast Pass and Translink as
of 06/16. Just like the TransLink cards, Clippers are accepted on Muni,
BART, AC Transit, Caltr |
|
w*******y
发帖数: 60932 | 38 First, I dislike the ThankYou network because they stopped issuing $100 gift
cards at useful places (e.g. Sears, Home Depot, Lowes), but I have three
Citi credit cards linked to this rewards program. So every now and then I
look to see what they have available.
Anyhow, although you can't get the $100 gift cards at Lowe's anymore, they
currently have a sale where you can get a $50 gift card for 5000 points (
normally 6000). In essence, you can get $100 worth of gift cards for 10,000
points, whi |
|
w*******y
发帖数: 60932 | 39 Here's the product link: Link:
http://gorillagadgets.com/bamboo-ipad-case.html
1. This is not a repost
2. Price Comparison tool showed nothing for this product: I then checked
google products(froogle.com) and found a similiar one on amazon.com for $29.
99 - but you DO pay for shipping. Google Products Link:
http://www.google.com/products?q=ipad bamboo case&spell=1&oi=spell
3. The price is $27.99, normally on the site for $34.99 - you get free
shipping always tho, so the shipping portion is |
|
w*******y
发帖数: 60932 | 40 Softpedia has a special WinX DVD Author free, full version download
available. The normal cost is $39.95. The registration code is included in
the download so no email registration is needed. These Softpedia promo's are
time limited, we just don't know how long. So if you want this be sure and
get it ASAP. You can find out more info about the promo here-
Link:
http://tricks-collections.com/winx-dvd-author-special-edition-full-version/
Here is the softpedia page-
Link:
http://www.softpedia.com |
|
|
w*******y
发帖数: 60932 | 42 Normally cost $7 a quart from Autozone and other auto builders. This is the
manufacturer recommended oil for Mercedes and BMW.
Amazon has this for $5.90 a quart- the cheapest I've seen yet in ages after
rebate.
$47.39 for pack of six 1 quart bottles - $12 rebate = $35.39
LINK:
Link:
http://www.amazon.com/gp/product/B000SKV4U2/ref=oss_product
Product Description
Mobil 1 0W-40 exceeding industry standards and the major leading builder
requirements is the cornerstone of the performance reserve that |
|
|
|
w*******y
发帖数: 60932 | 45 For those that don't know what this is, it's a berry that is native to West
Africa. When eaten, it temporarily alters your tastebuds. Sour foods taste
sweet, etc. You can bite into a lemon an it will taste like a sweet orange
! They're great for parties and a lot of fun to experiment with.
Read the reviews!
The price on Amazon is normally $17.99 for a ten pack, but it's only $10/10
berries right now. Amazon also offers a two pack for $25.00, but this deal
is even better than that one!
I was... 阅读全帖 |
|
w*******y
发帖数: 60932 | 46 jcpenney.com has a canopy swing for $59.49 (normally $260) after coupon with
free in-store shipping.
Use coupon code 2BMERRY or TOYTIME for the 15% off (both expire on 12/18).
Has good reviews and will be great to use when spring rolls around.
Link:
http://www5.jcpenney.com/jcp/X6.aspx?GrpTyp=PRD&ItemID=17a843c&Nao=0&hdnOnGo=true&Ne=840 4 877 878 5 961 6 29 3 598 11 15 12 506 10 23 585 596 1031 8 18 904 903 969 949 833&Ntt=canopy swing&SearchString=canopy swing&N=4294959029&SO=0&PSO=0&CmCatId=se... 阅读全帖 |
|
w*******y
发帖数: 60932 | 47 OnLive is that new online gaming system.
Comes with pass to get free game of choice.
Normally $99, now $66 though Sunday (CES Sale - no doubt related to their
announcement with Vizio).
Link:
http://www.onlive.com/game-system
|
|
w*******y
发帖数: 60932 | 48 This is my first deal post so go easy on me if I am not expert level..
Apple launched it's Mac App Store this week for apps for the computer, and
they have Apeture 3 for $79.99 and I guess it is normally like a $199 for
this photo editing program. It did not say when or if the price would be
going back up.
I did have to update my macbook software to 10.6.6 to get the Mac App Store.
.. App.
Happy Photo Editing.
|
|
w*******y
发帖数: 60932 | 49 Link:
http://www.sonycreativesoftware.com/upgrade/moviestudiotopro?di
I'm sure this won't hit FP status because it's a niche product but thought
it was a good deal and wanted to pass along.
The upgrade is normally in the $400 range for Sony Movie Studio users (which
is basically a Vegas Lite package albeit still very good), the above price
is only $5 more than Vegas Pro users pay for an upgrade. I've been a Vegas
user for several years and it's wonderfully easy to use as full-featured
editing so... 阅读全帖 |
|
|
