2008 First Semester STATS课程官方问答贴

mayigo

A group of biologists is interested in surveying the abundance of barrel cactus in a particular desert. They divide the desert area up into 1200 equal sized rectangular areas. Of these, 247 areas are too difficult to access, so are excluded. Of the remaining 953 areas, 50 are chosen at random to be surveyed. The biologists then randomly select a 100 square metre section of each of these areas to survey. The number of barrel cacti in each of the chosen sections is then measured.
1.is non-response bias a potential problem with the survey?briefly justify your answer.
2.is selection bias a a potential problem ?briefly justify your answer.
请问高手，我可不可以把247 areas 看作是non-response bias ？
而953 areas看作是整个1200areas看作是selection bias
另外如果一个调查有80%人回答和20%人不回答，那那个20%人不回答是non-response bias ，用80%人回答来代表整体是selection bias 吗？到底non-response bias 和selection bias 有什么区别？

z-score

原帖由 mayigo 于 2008-3-9 22:26 发表
A group of biologists is interested in surveying the abundance of barrel cactus in a particular desert. They divide the desert area up into 1200 equal sized rectangular areas. Of these, 247 areas ...

”而953 areas看作是整个1200areas看作是selection bias“

恩，这个你说得很对。Those that are too difficult to access may be totally different from the 953 that can be accessed.

"我可不可以把247 areas 看作是non-response bias ？"

他问的这个问题很tricky。Non-response bias是在你take一个sample时候没有得到回复所造成的。在这里，你take的sample size是50，50块土地你都得到了cactus number。所有non-response bias is not present。

Selection bias是difference between your sample and the population of interest。

Non-response bias是在你sample的那些人或东西里没有给你回复所造成的bias。

mayigo

原帖由 z-score 于 2008-3-9 22:41 发表


”而953 areas看作是整个1200areas看作是selection bias“

恩，这个你说得很对。Those that are too difficult to access may be totally different from the 953 that can be accessed.

"我可不可以 ...

谢谢，那lz,那如果953 areas是selection bias的话，，最后一个问题问这个是不是sampling errors,而selection bias是non-sampling errors,那这个survery 是否应该是non-sampling erro人?而且我也没有觉得他存在任何behavioural consideration,感觉这个survey 完全不是调查性质的，好像和non-response bias ，behavioural consideration没有任何关系似的，我都糊涂了，

[ 本帖最后由 mayigo 于 2008-3-9 22:54 编辑 ]

z-score

原帖由 mayigo 于 2008-3-9 22:52 发表

谢谢，那lz,那如果953 areas是selection bias的话，，最后一个问题问这个是不是sampling errors,而selection bias是non-sampling errors,那这个survery 是否应该是non-sampling erro人?而且我也没有觉得他存在 ...

恩，你不要被出题的人搞糊涂了。

一般不涉及到人类的survey不会有很多non-sampling error, 例如behavioural considerations, non-response bias, self selection bias, question and format effect都不会有。

Sampling error is always present in surveys. Without sampling error we will be doing maths, not stats.

mayigo

原帖由 z-score 于 2008-3-9 23:01 发表


恩，你不要被出题的人搞糊涂了。

一般不涉及到人类的survey不会有很多non-sampling error, 例如behavioural considerations, non-response bias, self selection bias, question and format effect都不会 ...

但是LZ不是认为953 AREAS是SELECTION BIAS的吗？那SELECTION BIAS是NON-SAMPLING ERROR（COURSE BOOK ）有说道，那这个SURVEY为什么不是non-sampling error？

z-score

原帖由 mayigo 于 2008-3-9 23:07 发表

但是LZ不是认为953 AREAS是SELECTION BIAS的吗？那SELECTION BIAS是NON-SAMPLING ERROR（COURSE BOOK ）有说道，那这个SURVEY为什么不是non-sampling error？

是的，这个survey有selection bias, which is a non-sampling error。还有sampling error。

mayigo

原帖由 z-score 于 2008-3-9 23:11 发表


是的，这个survey有non-sampling error。还有sampling error。

谢谢LZ，那就是说一个SURVEY，即可以有non-sampling error 也可以有sampling error

z-score

原帖由 mayigo 于 2008-3-9 23:15 发表

谢谢LZ，那就是说一个SURVEY，即可以有non-sampling error 也可以有sampling error

是的，你找到了精髓。任何统计study都会有non-sampling error和sampling error。我们的任务是去减小non-sampling by designing and executing the study more carefully。我们减小sampling error的方法是take a larger sample size。

mayigo

原帖由 z-score 于 2008-3-9 23:20 发表


是的，你找到了精髓。任何统计study都会有non-sampling error和sampling error。我们的任务是去减小non-sampling by designing and executing the study more carefully。我们减小sampling error的方法是ta ...

太谢谢LZ

glasszon

Come on ATM, where's all the resources for stats326?

mayigo

Study 1: A marketing manager for a chain of 64 small bakeries was interested in how different special offers affected sales. 20 of the bakeries were randomly chosen to have a special offer where customers could purchase three loaves of bread for the price of two. 20 of the remaining bakeries were randomly selected to sell loaves of bread with a 20% discount. The remaining bakeries did not offer any discount. After a week, the total number of loaves of bread sold at each bakery was recorded.
Study 2: A university researcher was interested in whether there is a difference, between commerce and arts graduates, in the number of different jobs students went on to work in after graduating. Random samples of 200 commerce and 200 arts graduates were drawn from the graduation lists in 1997. Each graduate was contacted and the number of different jobs they had worked in the 10 years since graduating was recorded.
Study 3: A road safety researcher was interested in the effects of alcohol on the driving ability of young adults. A sample of 50 young adults was obtained and each was randomly allocated to either an alcohol group or a no alcohol group. The members of the alcohol group each drank 700ml of beer then did a 10 minute session on a driving simulator. The members of the no alcohol group just did a 10 minute session on a driving simulator. For each member, the driving simulator reported the number of errors made during their 10 minute session.

Which of the studies, if any, included a control group? For the studies that include one, briefly describe the control group.
Which of the studies, if any, used placebos?
请问楼主，什么是control group，在这3个调查里有control group.
吗？
另外第二个问题怎么看似和题目无关，是不是药物的调查报告才有placebo

z-score

原帖由 mayigo 于 2008-3-11 01:55 发表
Study 1: A marketing manager for a chain of 64 small bakeries was interested in how different special offers affected sales. 20 of the bakeries were randomly chosen to have a special offer where  ...

你问了一个很好的问题。

在一个study里，我们可以把不同的group看为treatment group, placebo group和control group。一个study的最终结论是根据treatment group和control group的比较。例如Study 1, 你想知道的是discount对销量有没有影响。这里，你要比较买三送一，打八折和没有discount的店。如果买三送一和打八折店的销量比没有discount的高，那么我们就可以得出结论。

所以，如果在一个study里，你要compare with something, 那么那个something就是control group.

Placebo一般是在于吃药有关的study里。例如，吃药的group是treatment group, 不吃药的是control group, 吃糖药丸的是placebo, 因为人吃任何药都会有心理和生理作用，感觉自我良好(placebo effect).

原帖由 glasszon 于 2008-3-10 13:23 发表
Come on ATM, where's all the resources for stats326?

I don't have any.

[ 本帖最后由 z-score 于 2008-3-11 08:48 编辑 ]

当时的理想

WHat  is your sta10X's work example for?  Is not the answers for ASSIGN 1 ,isn't it ??
  

z-score

原帖由 当时的理想 于 2008-3-11 09:18 发表
WHat  is your sta10X's work example for?  Is not the answers for ASSIGN 1 ,isn't it ??
  

Nope, it's not, hehe. It's my own example of how to set out a STATS10x assignment. Hope it helps.

[ 本帖最后由 z-score 于 2008-3-11 10:23 编辑 ]

阴雨天

非常非常感谢LZ~~~ 好人一个

天外飞猪

太谢谢lz了..
我1定要努力跟努力学习
=我做到assign不会还得请教lz呢..
请多多帮忙

mayigo

non-response can cause bias in surveys because non-respondents can behave differently from people who respond
请问楼主能解释以下这句话为什么是对的吗？特别是non-respondents can behave differently frompeople who respond 这句话怎么解释，non-respondents 本来就不做回答，为什么能 behave differently frompeople who respond？

z-score

原帖由 mayigo 于 2008-3-12 17:28 发表
non-response can cause bias in surveys because non-respondents can behave differently from people who respond
请问楼主能解释以下这句话为什么是对的吗？特别是non-respondents can behave differentl ...

恩，问得很好，这句话概念比较抽象。举一个例子，一个marketing research公司做一个民意调查，这家公司派员工在白天给1000户人家打电话，问他们在08年的大选会选哪个政党。在白天，能在家里休息的很多是退休的老人，他们比较喜欢投优先党。白天在上班的人们比较喜欢投国家党和工党。如果在一个1000人的sample, 只有500个人回答了，而且这500人中很多是老年人，那么这个民意调查的结果就会有non-response, 因为从没在家的500 respondents和500 non-respondents得出的结果会不一样。这就是non-respondents can behave differently from people who do.

mayigo

[楼主太强了，回答的正是好，谢谢谢谢谢谢谢谢谢谢谢谢，哦，因为从没在家的500 respondents和500 non-respondents得出的结果会不一样。这就是non-respondents can behave differently from people who do.是不是应该因为从没在家的500 non-respondents和500respondents得出的结果会不一样。这就是non-respondents can behave differently from people who do.



[ 本帖最后由 mayigo 于 2008-3-12 22:53 编辑 ]

雨中小刀

有个问题啊..现在在做208的assignment 1, 第一题,在输入了survey.df<-read.table(file choose(),header=T,选择了文件的路径,从买的光盘里直接到s2008a,再回车以后出现
> library(s20x)
> survey.df<-read.table(file.choose(),header=T)
> attach(survey.df)

        The following object(s) are masked from package:datasets :

         sleep

> survey.df(1:5)
错误: 没有"survey.df"这个函数

数据就显示不出来,是怎么回事啊? 多谢啦...

z-score

原帖由 雨中小刀 于 2008-3-13 00:30 发表
有个问题啊..现在在做208的assignment 1, 第一题,在输入了survey.df library(s20x)
> survey.df attach(survey.df)

        The following object(s) are masked from package:datasets :

         slee ...

应该是
> survey.df(1:5, )

你少打了一个逗号。

z-score

原帖由 mayigo 于 2008-3-12 22:51 发表
[楼主太强了，回答的正是好，谢谢谢谢谢谢谢谢谢谢谢谢，哦，因为从没在家的500 respondents和500 non-respondents得出的结果会不一样。这就是non-respondents can behave differently from people who do.是不 ...

对的，现在你彻底理解了。

spirithell

Question 3. [12 marks] [Chapter 1]
Consider the following studies:
Study 1: A marketing manager for a chain of 64 small bakeries was interested in how different special offers affected sales. 20 of the bakeries were randomly chosen to have a special offer where customers could purchase three loaves of bread for the price of two. 20 of the remaining bakeries were randomly selected to sell loaves of bread with a 20% discount. The remaining bakeries did not offer any discount. After a week, the total number of loaves of bread sold at each bakery was recorded.
Study 2: A university researcher was interested in whether there is a difference, between commerce and arts graduates, in the number of different jobs students went on to work in after graduating. Random samples of 200 commerce and 200 arts graduates were drawn from the graduation lists in 1997. Each graduate was contacted and the number of different jobs they had worked in the 10 years since graduating was recorded.
Study 3: A road safety researcher was interested in the effects of alcohol on the driving ability of young adults. A sample of 50 young adults was obtained and each was randomly allocated to either an alcohol group or a no alcohol group. The members of the alcohol group each drank 700ml of beer then did a 10 minute session on a driving simulator. The members of the no alcohol group just did a 10 minute session on a driving simulator. For each member, the driving simulator reported the number of errors made during their 10 minute session.

(c) For the studies that are observational, could an experiment have been easily carried out instead? If so, briefly explain how. If not, briefly explain why not.
stuay 1 is experiment,这个就不讨论了。
study 2 虽然是observational reasearch,但是好像可以搞成experiment，请问怎么下手呢？
study 3 好像没法搞成experiment.
以上是我对Q3--(C) 的看法，请您帮我看看，怎么做这一题？谢谢谢谢。！~~~~

我现在就Q3和Q5不会做，唔唔唔唔唔唔~~~~一点点地问，谢谢阿童木。

spirithell

还是Q3， (d)已经看过上面的帖子了，
(e) It was later suggested that gender could be a significant factor in Study 3. How could the study design be modified to take into account the gender effect?能不能帮忙提示一下？
(f) Which of the studies, if any, included a control group? For the studies that include one, briefly describe the control group。 应该是study 3,自己胡编一下就行了。
(h) Could the results of a study such as Study 2 be used to argue that any overall difference in the number of different jobs held is due to the degree studied? Briefly justify your answer.实在是不明白，怎么还可以argue捏，晕晕滴说， ，不好意识，实在太笨了。

spirithell

Statistics Assistance Room

Assistance room tutors provide one-on-one help with work in lectures, assignments, and with test & exam preparation. You can also ask for help with the calculator worksheet.

When: Staffed from 9am – 4pm, Monday – Friday.

Where: Room B09, Basement of Building 303, Science Centre.

当时的理想

  LZ。。那些概念是理解了。。可是语言不知道怎么组织哦。。  就是无处下手。。知识点是懂了 。。。 只怪英语不那么好。。怎么办哦？

还有。。 USE THE TABLE OF RANDOM NUMBERS。。 那个不知道要怎么表达哦、、、 老师课上讲用自己手机号码的后2个数字来选一组号码出来。。。 ASSIG的题目也要那样吗？ 还是自己随便选哦 ？？？？

谢谢。。。。

z-score

study 2 虽然是observational reasearch,但是好像可以搞成experiment，请问怎么下手呢？

如果study 2变成experiment的话，学生们会被randomly allocated into the commerce or art faculty. 你觉得这样可能吗？可能强制一个人去读BCom或BA?

study 3 好像没法搞成experiment.

study 3 ms是experiment哦。你看这句话was randomly allocated to either an alcohol group or a no alcohol group。

我现在就Q3和Q5不会做，唔唔唔唔唔唔~~~~一点点地问，谢谢阿童木。

Q5可以参考4楼past assignment的答案。

[ 本帖最后由 z-score 于 2008-3-13 21:50 编辑 ]

z-score

原帖由 当时的理想 于 2008-3-13 21:18 发表
  LZ。。那些概念是理解了。。可是语言不知道怎么组织哦。。  就是无处下手。。知识点是懂了 。。。 只怪英语不那么好。。怎么办哦？

还有。。 USE THE TABLE OF RANDOM NUMBERS。。 那个不知道要怎么表 ...

哦，如果你怕自己的英文表达能力不好没事。你先写出来，我帮你改。如果你不想让别人看到你写的，你发给我PM也可以。

其实答案不用说很多，写一些简单的解释就好了。

[ 本帖最后由 z-score 于 2008-3-13 22:50 编辑 ]

z-score

原帖由 spirithell 于 2008-3-13 19:36 发表
还是Q3， (d)已经看过上面的帖子了，
(e) It was later suggested that gender could be a significant factor in Study 3. How could the study design be modified to take into account the gender effect? ...

(e) It was later suggested that gender could be a significant factor inStudy 3. How could the study design be modified to take into accountthe gender effect?能不能帮忙提示一下？

Q3 e) 很大的一个hint, blocking是什么东西？Blocking能在study 3用吗？

(h) Could the results of a study such as Study 2 be used to argue thatany overall difference in the number of different jobs held is due tothe degree studied? Briefly justify your answer.

A study can argue that any overall difference is due to the treatment group of the study if and only the study is a well designed and executed experiment. If the study is an observational study, then we cannot make such causation statements.