w****b
发帖数: 623 | 1 Guys, I still don't know why we are disputing on these issues. Or these
non-issues. Just leave the history to historians, as we will never see 1st
hand materials and we'll have to allow ourselves to be influenced by writers,
who, may well have biased opinions.
For the given hands, I find Truscott's accusation soundless. But then, for
cheating cases, typically each individual hand does not supply enough
evidence, as everybody has deviated from the routine once in a while. But
collectively, deviat |
|
l*******z
发帖数: 4276 | 2 You have said that after Panthers, after Texans, after Titans, etc.
There is a norm, and then there is a deviation from the norm. Even in an
extreme deviation like today, good team still finds way to win. That
happened to Den, to NE, to CIN, to NO, to every good team. And this type of
extreme does not happen often. Odds are still on our side. |
|
r****y
发帖数: 26819 | 3 摘自wiki：
http://en.wikipedia.org/wiki/First-move_advantage_in_chess
In chess tournaments and matches, the frequency with which each player recei
ves White and Black is an important consideration. In matches, the players'
colors in the first game are determined by drawing lots, and alternated ther
eafter.[162] In round robin tournaments with an odd number of players, each
player receives an equal number of Whites and Blacks; with an even number of
players, each receives one extra White or Black. W... 阅读全帖 |
|
h*********r
发帖数: 1943 | 4 http://www.snipercountry.com/InReviews/FNHUSA_SPR300WSM.asp
这篇文章讲了测试300WSM，其中有一段说膛口速度的标准差，300WSM只有6fps，而Fed
的GMM 300WM高达25fps：
The standard deviation of 6 fps is pretty incredible. I fired my 300WM with
190 GM300M side-by-side, same day, and the standard deviation was 25 fps.
That is about as "match" grade as you get.
据我所知应该是初速越稳定越准，弹壳太长的时候，装药点燃的一致性一般不如Short
Action，所以很多优秀Benchrest都是往胖弹壳方向发展的。
300WM要做到比308准，个人觉得需要更高级别的机械加工和匹配才有可能 |
|
c******i
发帖数: 4091 | 5 The Knox Report
From the Firearms Coalition
Kennesaw Revisited
By Jeff Knox
(October 7, 2008) In 1981, Morton Grove, Illinois became the first town in
the U.S. to pass a flat out ban on the possession of handguns within the
town limits by anyone except police and active duty military during the
performance of their official duties. In response, the town of Kennesaw,
Georgia passed a gun law of their own in March of 1982. The Kennesaw law
was almost the exact opposite to the Morton Grove ordina... 阅读全帖 |
|
K****D
发帖数: 30533 | 6 It's like comparing
A) average score 3:6, with standard deviation of 1.5 games.
B) average score 1.5:6, with standard deviation of 3 games.
的确，搏杀了确实有一线获胜的可能性，但是你和这个强手打100场，肯定是
稳扎稳打获胜的总局数多。
偶个人肯定是倾向于稳扎稳打，因为输得很惨对偶的打击比较大。偶的观点
是对手比偶强的情况下，偶本来就不应该赢，只有能够把比分咬接近后才有
可能通过实力的稍微增加获胜。通过搏杀的获胜只是一时的快活，因为没法
reproduce.
换句话说，如果偶和高手打球20场，赢了一场7:6, 其余19场不是bagel就是
stick, 偶的信心丝毫不会增加，不会认为“偶还是有戏可以赢这个家伙的”。 |
|
p*****l
发帖数: 399 | 7 I read about a formula in a book today, which calculates the probablity of
going broke.
Pr(Going broke)=e的[(-2BW)/(V)]次方=EXP((-2BW)/(V))
B is the bankroll
W is the win rate
V is the variance which is equal to the standard deviation squared
For example, for a good player in a good live game, a typical win rate is 1
big bet per hour and a typical standard deviation is 10 big bets per hour,
based on the assmuption, we can calculate the risk of ruin for various size
bankrolls as follows:
Bankroll (B |
|
W********m
发帖数: 7793 | 8 Pot size has no effect on whether your ev is + or - for one hand or over
long term. Variance also has no impact on true EV. Your actual money gain
or loss is only an estimate of the true EV. Variance has an impact on how
often your estimate can deviate from your true EV. It can deviate up or down
. The larger the sample size is, the less variance it has for the estimate
and it should always converge to your true EV. (We say variance, it is
really standard error which should be used to describe... 阅读全帖 |
|
T********n
发帖数: 528 | 9 His conclusion is wrong. Also you can still find good double deck games in
Vegas all over the place, the minimum is just higher.
Player said bankroll management is probably the issue. That's probably
right, if you give him the benefit of the doubt.
I don't - the fact that he uses hi-low split as his counting strategy of
choice indicates he's probably a beginner at card counting. Can't say for
sure since I didn't read his book.
hi-low split does a decent enough job with the count itself, but i... 阅读全帖 |
|
s*****s
发帖数: 1130 | 10 这个网站很不错,你知道怎么在HM2上查winrate的standard deviation么?
Assuming standard deviation of 100BB/100, 我这个大概是1%的概率,也不算太离谱. |
|
p******a
发帖数: 975 | 11 要注意的是这个网站的BB其实是bb，就是一个big blind。standard deviation的话一
般6max是100bb/100。但是这个也和打法有关，从你的描述看可能你的standard
deviation会更大一点。
你的例子确实比较极端，所以我建议你贴到大一点的论坛去问问。
还有一个问题是，我们能不能假设EV线就代表了真实的win rate。这个可能也是一个比
较大的假设，因为run的好的时候的EV线和run的不好的时候的EV线也会差很多。
谱. |
|
i*******8
发帖数: 7955 | 12 对，只有第一颗子弹是基本打到准星位置的，deviation每把枪都不一样，附件也改变
deviation |
|
l**********o
发帖数: 9952 | 13 我倒没有，但确实有很年轻的人。既然是Population，总有mean和standard deviation
嘛，两，三个，甚至三，四个standard deviation之外的个例总存在。 |
|
c****0
发帖数: 14490 | 14 对我的中文很捉急。。。
介是population standard deviation or sample standard deviation?
FOR |
|
s***e
发帖数: 5242 | 15 【 以下文字转载自 Faculty 讨论区 】
发信人: positiveforc (positive), 信区: Faculty
标 题: Re: 找AP职位的个人总结 － 四十二万公里路云和月
发信站: BBS 未名空间站 (Mon Mar 13 00:14:25 2017, 美东)
十一、我是歌手
在悉尼呆了不到两周，再次飞美国，这次去佐治亚的一个R1。这是一次毕生罕见的面试
、奇遇。面试从第一分钟开始就发生了我意想不到的变化，过程出乎意料。 面试前的
准备不说了，直接跳到面试开始的一刻。我和SCC次早饭，SCC是一个白人老头，小胡子
留着非常帅气。我们边吃边聊，他有一搭没一搭的说着，不热情也不冷淡，也就是很职
业化。他问我，你做的科研和音乐相关，你是musician 吗？我说不算吧，但是我爱听
歌唱歌。他问，你有最爱的歌吗？我其实喜欢的歌挺多的，我想既然来了佐治亚，就说
一首佐治亚的歌吧。我说我最爱Georgia on my mind，这是一首黑人歌，也是佐治亚的
州歌。这时我意想不到的事情发生了，我突然看到他的眼睛里放出了光芒，笑从两颊生
。我忘不了他的脸色从那一秒开始的秒... 阅读全帖 |
|
f******t
发帖数: 7283 | 16 现在咱们这行里早就有deviated well的成熟技术了。地上有不能破坏的东西，不怕，
咱们从远处开挖，井的走向逐渐变化，最后到达target zone。地上东西可以完美不受
影响。
Avatar那都是21xx年了，你要采矿石，用回一百多年前就有的deviated well技术就可
以了嘛，为何非要把人家的树给砍掉呢？几亿刀成本的电影剧情起码不要这么容易有不
make sense的常识愚钝啊。 |
|
b******g
发帖数: 4676 | 17 不客气，只是想确认你在做客观的比较。其实你想比较什么，我都是没意见的，但是只
希望
你要比较的时候一定要客观，不能简单的凭感觉就来下结论，这个是必要的，也是你经
常犯的一个错误。下面谈一下我的理解，不对的话大师们来批评指正。
如果想比较噪点，需要有一个客观的标准。noise，一个常用的衡量指标是standard
deviation。但是同样std的noise，给人的感觉可能会是不一样的，因为你看到signal
的强度不同。因此图像里面通常更关心的是SNR，也就是signal to noise ratio, 常用
定义是: signal mean/noise standard deviation.
啰嗦这么多，是想告诉你，如果你想看噪点，那么找一个最黑最黑的没有任何信号的区
域来比较噪点才是客观的。如果你想比较噪点在你的图片里面是否严重，你是否可以接
受，你需要确保比较的信号是一样强的才有意义。也就是说，rose给你个大白天拍出来
的iso 1000，和你半夜拍出来的iso 1000，噪点虽然一样，但是感觉可能不会是一样的。
想比较噪点，看看那些专业网站是如何做测评的，按照人家的方法测一... 阅读全帖 |
|
G**Y
发帖数: 33224 | 18 Patent of the image plane phase difference AF not to degrade the image
quality Canon [semiconductor electronics technology] [edit]
キヤノンが像面位相差AFを用いても、画質が劣化しない撮像素子に関する特許を出願
中です。 Canon patents pending on the imaging device may be used to image
plane phase difference AF, image quality does not deteriorate. この技術なら
、既存のレンズ資産の利用と高速AFを可能としつつも画質の劣化がほぼありませんか
ら、後発であっても他社に負けないミラーレスを出せますね。 If this technology,
since almost no degradation of image quality while still enabling high-
speed AF lens and the use of... 阅读全帖 |
|
n**b
发帖数: 13203 | 19 【 以下文字转载自 SanFrancisco 讨论区 】
发信人: qdssl (honeydot), 信区: SanFrancisco
标 题: FDA查处美国海尔斯制药有限公司在中国热卖的一氧化氮等保健产品 (转载)
发信站: BBS 未名空间站 (Sat May 5 13:29:53 2012, 美东)
发信人: qdssl (honeydot), 信区: LosAngeles
标 题: FDA查处美国海尔斯制药有限公司在中国热卖的一氧化氮等保健产品
发信站: BBS 未名空间站 (Sat May 5 13:27:59 2012, 美东)
Health One Pharmaceuticals, Inc. of El Monte, CA was warned by the FDA that
a September/October inspection of the company's dietary supplement
manufacturing facility revealed serious violations of FDA's Current Good
Manu... 阅读全帖 |
|
e**n
发帖数: 5876 | 20 对焦精度非常重要, 但做对焦精度测试的网站不多, optyczne是其中一个.
测试方法是, 照40张(a series of 40 images), 然后MTF50分布来判断对焦精确度(The
results are presented in the form of a histogram, which represents the
percentage of deviation from the best MTF50 measurement in the series)
在40张测试照片中, A6000对焦精准的(MTF在deviation 1.5之内)的就是只有11张. GX7
有28张是精准的, EM10最好, 29张, NX30是20张.
optyczne也指出了, 镜头会对测试结果有影响, 开始a6000用的第一个24f1.8,对焦精确
度更糟糕.无奈只好有换了一个copy. 上图是第二个copy的结果
同样的测试在tungsten光源下重复了一遍
差距还是巨大 |
|
e**n
发帖数: 5876 | 21 A6000曾经追焦领先, 但后来被用DFD技术的GH4超过, DFD会用在P+全线相机, 比如新出
的G7
A6000的单次对焦从来就比P+的差. 无论是速度还是精度还是暗光对焦能力. optyczne
做个对比测试:
测试方法是, 照40张(a series of 40 images), 然后MTF50分布来判断对焦精确度(The
results are presented in the form of a histogram, which represents the
percentage of deviation from the best MTF50 measurement in the series)
在40张测试照片中, A6000对焦精准的(MTF在deviation 1.5之内)的就是只有11张. GX7
有28 out of 40是精准的, EM10最好, 29张精准, NX30是20张.
optyczne也指出了, 镜头会对测试结果有影响, 开始a6000用的一个24f1.8,对焦精确度
更糟糕.无奈只好有换了一个copy. 上面的结果是第二个copy的结果
同样的... 阅读全帖 |
|
e**n
发帖数: 5876 | 22 E5, EM5....
另外也可以考虑gx7. 板上刚有人问过em10 vs gx7的选择
m43的镜头可玩的多了, 定焦推荐:
7.5f3.5(备选8f3.5, 8f1.8)
12f2
15f1.7(备选14f2.5, 17f1.8,17.5f0.95)
20f1.7
25f1.4(备选25f1.7, 25f1.8,25f0.95)
42.5f1.7(备选45f1.8, 42.5f1.2, 42.5f0.95)
75f1.8
还有3个微距头
变焦３大元, 和其他各类变焦, 看你需求了.
a6000的芯片大一点, 但是很多镜头光圈小, 比如受s+用户推崇的simga三个f2.8定焦
. f2.8,还没防抖, 谈什么高感优势 再比如恒定f4, m43是恒定f2.8
A6000的对焦就追焦快, 但现在被用DFD技术的新P+机身超越了. a6000单次对焦不
如m43快, 而且精度很差. 而且暗光对焦能力比P+差远了.
这是一个测试, 测试方法是, 照40张(a series of 40 images), 然后MTF50分布来判
断对焦精确度(The results are prese... 阅读全帖 |
|
e**n
发帖数: 5876 | 23 另外, 对焦精度非常重要, 对焦精度差的更耽误事, 还不不如慢点呢.
做对焦精度测试的网站不多, optyczne是其中一个.
测试方法是, 照40张(a series of 40 images), 然后MTF50分布来判断对焦精确度(The
results are presented in the form of a histogram, which represents the
percentage of deviation from the best MTF50 measurement in the series)
在40张测试照片中, A6000对焦精准的(MTF在deviation 1.5之内)的就是只有11张. GX7
有28张是精准的, EM10最好, 29张, NX30是20张.
optyczne也指出了, 镜头会对测试结果有影响, 开始a6000用的第一个24f1.8,对焦精确
度更糟糕.无奈只好有换了一个copy. 上图是第二个copy的结果
同样的测试在tungsten光源下重复了一遍
差距还是巨大 |
|
g****g
发帖数: 1828 | 24 标准差（Standard Deviation），在概率统计中最常使用作为统计分布程度（
statistical dispersion）上的测量。标准差定义为方差的算术平方根，反映组内个体
间的离散程度。测量到分布程度的结果，原则上具有两种性质：
1. 为非负数值，
2. 与测量资料具有相同单位。
一个总量的标准差或一个随机变量的标准差，及一个子集合样品数的标准差之间，有所
差别。其公式如下所列。
标准差的观念是由卡尔·皮尔逊 (Karl Pearson)引入到统计中。
目录
[隐藏]
* 1 阐述及应用
* 2 标准差的定义及简易计算公式
o 2.1 标准计算公式
o 2.2 简化计算公式
o 2.3 随机变量的标准差计算公式
o 2.4 样本标准差
o 2.5 连续随机变量的标准差计算公式
o 2.6 标准差的性质
* 3 范例
* 4 正态分布的规则
* 5 标准差与平均值之间的关系
* 6 几何学解释
... 阅读全帖 |
|
d********1
发帖数: 3828 | 25 总有人说小概率事件，基数大了也会发生。
那我问，为什么出一个短跑天才能100米跑八秒这样的小概率事件，在70多亿人这么大
的基数上都没发生呢？
假设70亿人每天都跑10次100米，跑10年，那就是25000000000000个样本。这么大的基
数会不会出现8秒这样的小概率事件呢？
有些人不明白概率上有一种distribution叫做“short tale”。比如normal
distribution就是short tale。在2个standard deviation之外还有5%概率。但是4个，
6个，8个standard deviation之外的概率就几乎没有了。
不要以为人口基数有多大。数学上概率趋于零可以是指数级的，不要说人口基数，就是
你用全宇宙的粒子数做基数恐怕都没用。 |
|
z*********e
发帖数: 10149 | 26 http://www.weidai.com/smart-losers.txt
Given that there is significant existing variation in human intelligence,
it's curious that we are not all much smarter than we actually are.
Besides the well-known costs of higher intelligence (e.g., more energy
use, bigger heads causing more difficult births), it seems that being
smart can be a disadvantage when playing some non-zero-sum games. Here is
one example.
Consider an infinitely repeated game with 2n players, where in each round
all players are r... 阅读全帖 |
|
z**********e
发帖数: 22064 | 27 【 以下文字转载自 Military 讨论区 】
发信人: zhonghangyue (中行说), 信区: Military
标 题: 哈萨克斯坦的纪念雕像
发信站: BBS 未名空间站 (Tue Jan 27 11:36:09 2015, 美东)
好像两个侏儒在玩自拍
http://www.bbc.com/news/blogs-news-from-elsewhere-28016580
25 June 2014 Last updated at 06:05 ET
Kazakhstan: 'Selfie' statue removed after ridicule
A monument in Kazakhstan has been hastily removed after locals likened the
figures on it to hobbits taking a "selfie", reports say.
The statue in the city of Ust-Kamenogorsk was built to honour two 19th
Century figu... 阅读全帖 |
|
C*******g
发帖数: 9288 | 28 Newfoundland, India, Iran, Afghanistan, Venezuela, Burma, the Marquesas, as
well as parts of Australia use half-hour deviations from standard time, and
some nations, such as Nepal, and some provinces, such as the Chatham Islands
, use quarter-hour deviations.
记不清了，好像有 |
|
C*****e
发帖数: 367 | 29 【 以下文字转载自 Church 俱乐部 】
发信人: CCBible (神同在圣经), 信区: Church
标 题: Glossary to the Westminster Confession
发信站: BBS 未名空间站 (Tue Apr 24 21:16:33 2012, 美东)
Glossary to the Westminster Confession
Explanations of all archaic and uncommon words, usages, and expressions, in
the order in which they occur.
Title Confession. n. Public statement of religious beliefs.
1.1 Providence. n. Divine care. Manifest. v. Show. Are they. idiom. T
hey are. Unto. prep. With regard to. Sundry. adj. Separate. D... 阅读全帖 |
|
d******8
发帖数: 2191 | 30 In my lab manual, uncertainty is defined as standard deviation over Sqrt[n],
where n is number of measurements and standard deviation is Sqrt[ Sum[(x_i-
Ave(X))^2] /n ].
So uncertainty=Sqrt[Sum[(x_i-Ave(x))^2]] /n.
Does anyone know this?
I did calculation on coins and found that, according to that definition,
uncertainty= Sqrt[ (0-1/2)^2*n/2+(1-1/2)^2*n/2]/n -> 0 if n->infinity.
( I assume heads and tails both numbering n/2.)
uncertainty is ZERO! No uncertainty???????
Usually I treat uncertain |
|
a***n
发帖数: 404 | 31 请教一种统计量，希望能够描述下列的数据特征：
比如从一段时间采集来的数据，有些数据是集中在一段时间内出现的，有些数据是长时
间均匀分布的。现在要一个统计量区分他们（不同于standard deviation,见下面）
举个例子：
walmart卖的牙膏的日销售量就是常年基本均匀分布的，但是圣诞树就一般是集中在12月
左右这个时间段销售量比较大。
什么统计量比较适合表征这两种数据的这种特性呢？就是说数据的locality特性。注意
，不一定是集中在一个范围内，如果在多个离散的范围内有很强的集中性，这个统计量
要求也要有很强的表达，所以用standard deviation估计不行。比如某个商品夏天和冬
天销量都很强，其他时候没人买，这样的商品也要能区别于牙膏这样的均匀态的商品。
但是这个又好像走入了一个悖论，因为平均分布的数据，貌似也可以看作是由多个很陡
的均匀分布的“峰”构成的。。。俺也变得糊涂起来了，不过总觉得应该有个什么能够
系统的描述下区分这种数据特性的方法。
有啥统计量可以搞定这个么？有人说 entropy,我看不靠谱，至于clustering,还有其他的
很多复杂的数据分析方 |
|
g***l
发帖数: 18555 | 32 TSQL不是做统计用的，SQL SERVER 有这些
SELECT VAR(grandtotal_amount ) 'Variance',
STDEVP(grandtotal_amount ) 'Standard Deviation',
STDEV(grandtotal_amount ) 'Standard Deviation',
VARP(grandtotal_amount) 'Variance for the Population'
FROM ordr |
|
发帖数: 1 | 33 估计出题人想要的答案是 N = 100*sqrt(10/2)
但是这个题还是出的有问题。
（1） std deviation s = sqrt (sum(x(i)-u)^2/N) 是个 random variable, 因为每
个sample x（i）是 random variable.
（2） standard deviation的期望值/均值 是个constant： E(s) = E[sqrt (sum(x(i)
-u)^2/N) ] = C
（3）均值的估计值的误差也是个 random variable： ue= sum(x_i)/N - u； E（ue）
= 0; var(ue) = var(x)/N; |
|
R*****g
发帖数: 33 | 34 我也来讨论一下。那些光喊容易不给答案的机器人就不要来掺乎了. :-)
这是楼主改了几遍的原题，还是一团乱麻，
（我写的具体一些，是说100个用户给出评价，平均分是55，标准差是10，经理觉得可
信度不太高，想把标准差降低到2，然后问大概要多少用户。 我还画了图演示.
我同意wdong所说，如果样本是iid，不管样本size多大，样本标准差(sample standard
deviation， which is an estimation of standard deviation of population)是差
不多的。第一个标准差应该就是指这个。那第二个标准差就只能理解成样本平均值(x_
bar=sum(x1:xn)/n, which is a random variable)的标准差了。我知道从题目的表述
来看这么理解很牵强，但只能这样，要不然没题可做。
给定样本size 100, 样本平均值x_bar的标准差只有1 (population std which is 10,
divided by sqrt(100)).这其实很直观：随便抽一个，标准差大概10. 100个平均下来
... 阅读全帖 |
|
l****z
发帖数: 29846 | 35 这个, 不要被standard deviation吓到了,那个其实和题目正解没有什么关系.
假设那个平均是Y, standard deviation 是x,那么,
58 --------->Y---------------->98
58+ 2*x = Y Y+ 3*x =98
初中的二元一次方程而已. |
|
B*****i
发帖数: 1246 | 36 https://www.e-access.sbc.com/erecruit/sap(bD1lbiZjPTYwMCZwPTMxNTE3)/bc/bsp/
sap/hrrcf_pinstappl/application.do?BspClient=600&BspLanguage=EN&rcfSpID=0004
&rcfContext=ATT&PARAM=
UElOU1RfR1VJRD00OUJBRTQzMEZCRDkwOEZFRTEwMDAwMDA5NkVBNjM4Rg%3d%3d
到主页登陆career -> search position,在线投简历。
AT&T E-Recruiting
External Publication of Job Posting 52210692
Job Posting Title
Lead Tax Accountant
Published
03/18/2009
Reference Code
lead tax acct1 - web
Job Title
Lead Tax Accountant
Company
AT&T
Department
Finance
R... 阅读全帖 |
|
f**********t
发帖数: 32 | 37 In evaluating the resonableness of an entity's accounting estimates, an
auditor normally would be concerned about assumptions that are
1) susceptible to bias
2) consistent with prior periods
3) insensitive to variation
4) similar to industry guidelines
In evaluating the resonableness of an accounting estimates, an auditor most
likely would concentrated on key factors and assumptions that are
1) consistent with those of prior periods
2) similar to industry guidelines
3) objective and not suscepti... 阅读全帖 |
|
s****i
发帖数: 87 | 38 assessed control risk too low - 是说审计评估的控制风险过低了，这是贬义
so decreased substantive testing - 因为审计认为风险低，所以他们自然就减少了
实质性测试
上面这种情况，是在实际的deviation rate 大于审计测试的样本的deviation rate 时
才会发生。
我说点儿大白话，就是实际的错误多，但审计抽到的样本正好错误少，审计就认为内控
不错，减少了实质性测试。
还有一个对应的概念control risk too high,对比着理解就好多了！ |
|
s*****0
发帖数: 357 | 39 Essay写累了上来灌灌水，没想到脸红脖子粗的争了一场，多少有些意气了。忙了一天
，现在闲下心来好好说说这个话题。既然有网友抛砖引玉了，缄口不言也实在不够尊重
。另外版主塞了不少包子,也不能吃白食.
先说说常用的描述数据的几个概念.
1. random variation -- variability有两大类,一类是知道来源的,还有一类是
unexplained. 比如一部分variability是由treatment effect引起的,而剩下的大部分
variability却不知道缘故. 后者经常被归入random variation.
2. Mean -- 最常见的当然是arithmetic mean, 特殊情况下也会用geometric及
harmonic mean. 任何数据组都能计算mean, 但不是所有时候用mean都合适. 当碰到
extreme data values的时候(经常因为技术条件所限边缘值不太容易被精确测量),
median就会有很大的优越性, 因为median可以大幅降低个别extreme value的影响.
3. Median -- 概念不多提了, 优... 阅读全帖 |
|
y***i
发帖数: 11639 | 40 我的说法也不对
http://en.wikipedia.org/wiki/Standard_deviation
文章里找找这句话， 它下面的公式就是两个随机数和的标准方差计算公式。
The standard deviation of the sum of two random variables can be related
to their individual standard deviations and the covariance between them:
虽然要计算a-b 把b转成 -b，相减就变成相加：a+(-b)，就能用上面的公式了。 |
|
x*****a
发帖数: 972 | 41 做了一组ELISA，每个时间点得到3个数据，算了平均值和standard deviation。现在
reviewer让
把平均值换成log的，这没问题；但是standard deviation怎么做？
直接把原来的STD取log？ |
|
c******r
发帖数: 3778 | 42 应该可以做multiplex reaction
GAPDH是可以做所谓的internal control的。就是说，在同一个pcr中，同时run两个甚
至两个targets，其中一个是control，另外一个是要测的基因。
这个internal control的目的有两个，一个是quality control，就是保证这个
reaction没有忘掉任何东西，没有加错任何东西等等。另外一个目的是loading
control，assumption就是两个sample中的GAPDH是不变的，换句话说，要测量的是在同
样量的GAPDH下，要测的基因在处理前和处理后有无变化。
为了达到这个目的，最好是run multiplex。这样才能保证测到的GAPDH和G之间是来自
确定的同一个loading量。
另外一个run multiplex的原因是统计的原因。如果分开run GAPDH和G，那么
triplicate之后会得到三个G的量，和三个GAPDH的量，而这六个量之间没有一一对应的
关系。就是说不能用G1/GAPDH1，G2/GAPDH2。。。。而只能用G的平均值/GAPDH的平均
值。但是... 阅读全帖 |
|
c****n
发帖数: 21367 | 43 Holism是拿来理解复杂系统的，简单系统不需要，对吧？
复杂系统跟简单系统一样有均值有平均场有一切一阶的特性，
这些东西的统计手段我们已经学会且用得很多了。
难于刻画的，也就是我们搞不懂的地方，就是large deviation，
heavy tail or whatever you name it，这些fancy名词说的其实
都是同一个地方，从随机数学的角度来看，就是高阶矩的特性。
理解复杂系统的关键就在于掌握用高阶矩进行系统特性区分/建模的办法，
large deviation theory就是个入门尝试。当然还有其他别的尝试，
总体来说，目前进展不是很多。 |
|
s****n
发帖数: 234 | 44 Assuming you have done three independent data collections, each of which
contains dataset from 5000 molecules, I would analyze the data in two
different ways:
(1) calculate the statistics (mean velocity, mean FRET efficiency or ...)
for each 5000-molecule dataset, calculate the standard deviation of the mean
of the three statistical values (SEM), and present the data as
+ SEM;
Or
(2) pool all the data from three data collections and calculate the mean
statistics +/- Standard Error o... 阅读全帖 |
|
y***i
发帖数: 11639 | 45 【 以下文字转载自 Statistics 讨论区 】
发信人: yuuli (听，...听), 信区: Statistics
标 题: 请问诸位chi-square test
发信站: BBS 未名空间站 (Mon Jun 25 18:52:20 2012, 美东)
看文章
http://jcb.rupress.org/content/180/4/665.full.pdf+html?with-ds=
figure 2C 用了chi-square test 不能理解
我的理解数据是这样的：
以figure2 C,chromosome 2的图，第5天的数据为例子 （红框内数据）,这一天的数据
是monastrol significant,所以有个星号
control conditon有5%的cell 有deviation (supplimentary tableS2 红框内数据)
monastrol condition下， clone 10 有21.9%的cell 有deviation(supplimentary
tableS2 红框内数据)，然后根据chi square test,这是... 阅读全帖 |
|
c***y
发帖数: 615 | 46 Thank you very much for your explanation. I am reading some relevant book
now, slowly though.
I checked my log file and got the following informatuiion:
Average standard deviation of split frequencies = 0.075483
Maximum standard deviation of split frequencies = 0.345682
Is it good or not?
Many articles mentioned manual adjustment after alignment. Does it mean I
just visually check and then do the adjustment? It sounds very time-
consuming and hard to be error-free since I hav... 阅读全帖 |
|
G*****h
发帖数: 320 | 47 补充一下：
Grubbs' and ROUT tests 方法的前提是数据符合 Gaussian distribution，同时考虑
样本大小等。应该比单单考虑 SD 要更加合理。但不合适 不是Gaussian 分布的数据。
如果用 SD 倍数来判断，最近有人提出了一个替代方案，可以作参考：
"Detecting outliers: Do not use standard deviation around the mean, use
absolute deviation around the median" Journal of Experimental Social
Psychology, Volume 49, Issue 4, July 2013, Pages 764–766.
http://www.sciencedirect.com/science/article/pii/S0022103113000
纯属个人看法，仅供参考 :) |
|
N*********6
发帖数: 2302 | 48 【 以下文字转载自 Parenting 讨论区 】
发信人: ananpig (●○ 围棋数学一把抓的安安猪), 信区: Parenting
标 题: Truth Telling Is Academia's Privilege (and Obligation)
发信站: BBS 未名空间站 (Tue Jan 19 01:04:20 2016, 美东)
By Jay P. Greene
My advice for scholars wishing to contribute to public discourse during this
election season is simple: Just tell the truth. One might think it
unnecessary to urge scholars to be honest, but it is shocking how easily
education researchers are tempted to deviate from the truth in the hope of
gaining influ... 阅读全帖 |
|
w******g
发帖数: 67 | 49 Q1:
The definition of SFRation is (E(R)-L)/sigma, E(R)is the mean of
distribution of return and L is the threshold of your return. We you
calculate the probability when return equals threshold, you need the formula
z=(X-E(R))/sigma and put X=L. So you see there is a "-" between SFratio and
z value to calculate threshold.
Q2:
"standard deviation of sample means" means the standard deviation of the
distribution of the means of samples, which is also called standard error.
There are many ways to sa |
|
z***e
发帖数: 1757 | 50 The only pure strategy Nash is two of them located at 1/4
point, the other two located at 3/4 point. To show this is
indeed a Nash, you need show none of them has incentive to
deviate.
To show its uniqueness, just show any other location profile
is not a Nash, i.e at least one of them has incentive to
deviate. Be patient to discuss different cases one by one.
It's not hard. |
|
