2008 First Semester STATS课程官方问答贴

Vy2

so is the 14/50 rule related to the normality assumption?
like.. data>50 theres strong skewness.. therefore it does not match the normality assumption
thus when data>50 we shouldn't use t-test??
so by using the 14/50 rule.. what kind of conditions given from the rule tell us to use t-test or not?
god im so confused...

Vy2

  new question.. how do we calculate residual??
from memory the question was:
researchers investigating about height (as explanatory variable) and weight (as response variable)
y hat = -101 + 74x
now a female with height of 174cm and 60.57 kg is added
what is the residual??
i found in the course book, residual (u hat = y1 - y1 hat)
but.. have no idea how to calculate...

thx atm.. ^^

z-score

原帖由 Vy2 于 2008-6-8 23:09 发表
so is the 14/50 rule related to the normality assumption?
like.. data>50 theres strong skewness.. therefore it does not match the normality assumption
thus when data>50 we shouldn't use t-test? ...

Ok, I see where your confusion is. The normality assumption can be checked visually in STATS10x, which is examining the plots for skewness. When you decide if the data are skewed or not, you can apply the 14/50 rule. When you have decided the data are skewed, you may apply the Central Limit Theorem where if the sample size (n) is greater than 30, you can approximate for normality.

z-score

原帖由 Vy2 于 2008-6-9 01:24 发表
  new question.. how do we calculate residual??
from memory the question was:
researchers investigating about height (as explanatory variable) and weight (as response variable)
y hat = -1 ...

Yes you are absolutely right, the residual is actual(or observed) value - fitted(or predicted) value

To calculate the fitted(or predicted) value for the example above, you can put 174 in as x.

eg. -101 + 74*x ( I think the equation is different, cause I calculated it and the number was funny)

Once you have calculated the fitted, use 60.57 - the fitted which we calculated above.

阴雨天

06年  10X second semester exam的第9题:我怎样判断那个符号是大于等于还是小于等于 然后如果是 2 tailed test 整个式子就要*2对吧？ 谢谢！

ミ﹏龍.

阿童木哦，20x 2007SS 啲第8题，

选项2是怎么算出来的？

怎么预测-1年啲数据呢？........谢谢

还有一个问题，back transform时候，除了log可以用计算器算回来...

其他的transformation怎么用计算器back 回来呢？我只会算(eX-1)*100 这种~

拜托教一下！thx

[ 本帖最后由 ミ﹏龍. 于 2008-6-11 17:33 编辑 ]

阴雨天

06年 2nd semester exam的第15题. 这种题请给我讲解一下,拿到题一点思路都没有..

z-score

原帖由 阴雨天 于 2008-6-11 17:30 发表
06年  10X second semester exam的第9题:我怎样判断那个符号是大于等于还是小于等于 然后如果是 2 tailed test 整个式子就要*2对吧？ 谢谢！

Chi-square test和F-test都是one-sided upper tail test. 因为这两个distribution都不是symmetric.

z-score

原帖由 阴雨天 于 2008-6-11 18:13 发表
06年 2nd semester exam的第15题. 这种题请给我讲解一下,拿到题一点思路都没有..

Confidence intervals的assumptions在10x只要看normality就好了，那些standard deviations的回答都是distractors. 这样一下就能杀下三个选择。

从appendices的图看data基本symmetric, 有一些slightly skewed，所以confidence interval is valid.

z-score

原帖由 ミ﹏龍. 于 2008-6-11 17:31 发表
阿童木哦，20x 2007SS 啲第8题，

选项2是怎么算出来的？

怎么预测-1年啲数据呢？........谢谢

还有一个问题，back transform时候，除了log可以用计算器算回来...

其他的transformation怎么用计算器 ...

首先，208不讲其它的transformation, 知道一个log就ok了，coursebook里面其它的直接ignore就好了。

4th quarter 1989也不能算-1年的数据，不过能算是t = 0. 那么就是intercept加上4th quarter dummy variable coefficient然后back transform.

ミ﹏龍.

谢谢阿童木.....那4th quarter的dummy varible coeffecient怎么找？....

z-score

原帖由 ミ﹏龍. 于 2008-6-11 22:55 发表
谢谢阿童木.....那4th quarter的dummy varible coeffecient怎么找？....

It's in the output.

wangismee

How to do the following question,help plz
it is in 2nd Semester 2007 stats108 exam paper

7. The average lifetime of 75-watt incandescent light bulbs is 750 hours. The

lifetime (in hours), I, for such a bulb is modelled by a Normal distribution,

mean μ = 750 and standard deviation σ = 125.

In summary:

    F ∼ Normal(μ = 10, 000, σ = 1, 500) (compact fluorescent)

    I ∼ Normal(μ = 750, σ = 125) (incandescent)

Which one of the following statements is false?

(1) It is less likely for a 75-watt incandescent light bulb to have a lifetime

shorter than 550 hours than it is for a 20-watt compact fluorescent light

bulb to have a lifetime shorter than 7, 500 hours.

(2) About 84% of 20-watt compact fluorescent light bulbs have a lifetime longer

than 8, 500 hours.

(3) It would be unusual for a 75-watt incandescent light bulb to have a lifetime

shorter than 350 hours.

(4) About 2.5% of 75-watt incandescent light bulbs have a lifetime longer than

1, 000 hours.

(5) It is more likely for a 75-watt incandescent light bulb to have a lifetime

longer than 900 hours than it is for a 20-watt compact fluorescent light

bulb to have a lifetime longer than 12, 000 hours.

z-score

原帖由 wangismee 于 2008-6-12 10:19 发表
How to do the following question,help plz
it is in 2nd Semester 2007 stats108 exam paper
7. The average lifetime of 75-watt incandescent light bulbs is 750 hours. Thelifetime (in hours), I, for ...

你有graphics calculator吗？有的话就很好做．

z-score

原帖由 wangismee 于 2008-6-12 10:19 发表
How to do the following question,help plz
it is in 2nd Semester 2007 stats108 exam paper
7. The average lifetime of 75-watt incandescent light bulbs is 750 hours. Thelifetime (in hours), I, for ...

而且你可以看一下308,309楼的帖子，很相似．

~Kurosagi~

ATM:

Stats108

06年second semester的第32 題怎么算了? z

這一題和07年First Semeter的第44題是一樣的 但是都不會做

哈有就是 06年Second Semester的第28題, 用什麽方法計算的啊?

謝謝

z-score

原帖由 ~Kurosagi~ 于 2008-6-12 14:15 发表
ATM:

Stats108

06年second semester的第32 題怎么算了? z

這一題和07年First Semeter的第44題是一樣的 但是都不會做

哈有就是 06年Second Semester的第28題, 用什麽方法計算的啊?

謝謝

#32

residual = actual - predicted

actual = 140.21 in this case, the predicted is the value you get when putting 41 in the fitted equation.

#28

看confidence interval就好了．

Cagalli

ATM:
请问怎样看出scatter的分布图是否是constant scatter or non-constant scatter?
我听了老师的讲解，是以中线为标准，然后大约在左右两边划出一条线 然后来决定
但是无法掌握其中要领。。具体该怎么画
e.g  course book Chapter 12, pg17的examples的左边数下来第三还有第四张，为何第三张是no problem,第四张确是non- constant scatter?


另外请教regression 和correlation有什么区别呢？
谢谢！！~

z-score

原帖由 Cagalli 于 2008-6-12 15:23 发表
ATM:
请问怎样看出scatter的分布图是否是constant scatter or non-constant scatter?
我听了老师的讲解，是以中线为标准，然后大约在左右两边划出一条线 然后来决定
但是无法掌握其中要领。。具体该怎么画
...

这样最好，在mean画一条直线，然后在mean得直线上下各划一个平行线，然后把observations比较一下就好了．

Cagalli

好的，谢谢
请问regression 和correlation有什么区别呢？

阴雨天

请问在表示平均数的符号mu上面加以个短短的波浪线是什么意思呢？

Cagalli

应该是中位数把。。。
median

Cagalli

ROBUST是什么意思呢？会用在什么地方呢？
还有validity...谢谢

~Kurosagi~

thx~~~~~~~

阴雨天

原帖由 Cagalli 于 2008-6-12 16:13 发表
应该是中位数把。。。
median

谢谢啦。。。。。

阴雨天

2006 10X 2nd semester exam
Q27. 我根据什么claim the mean amount of TV watched per day by children between 2 country are different or same? 谢谢！

Vy2

原帖由 z-score 于 2008-6-9 08:38 发表


Ok, I see where your confusion is. The normality assumption can be checked visually in STATS10x, which is examining the plots for skewness. When you decide if the data are skewed or not, you  ...

couldn't log in for days don't know y.. anyways..
so according to the 15/40 rule.. n>40 may be strongly skewed..
that means we can not use t-test? since it doesn't match normality assumption....

and..
2-tailed are  more robust with respect to normality assumption han 1-tailed
that means.. when the 2-tailed test doesn't match the normality assumption.. it's ok.. right?

怎麽越讀越不懂.. @.@
thx atm,..

阴雨天

2006 2nd semester exam Q35 应该怎么算？ 带哪些数进去？

阴雨天

还是06年第一学期的卷子。
第三十六题 为什么是第四个选项呢？ 我感觉选第三个也挺合理的呀

阴雨天

2006 2nd semester exam
Q40为什么是situation c而不是 situation b呢？？？