[求助]这样的数据可不可以用panel data model处理

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

一组数据：

variety yield year place

A 30 1999 1

A 35 1999 2

A 32 2000 1

A 31 2000 2

B 30 1999 1

B 31 2000 1

B 33 2000 2

我定义variety为个体，year为时间。因为有不同的地域，所以用tsset的时候说有重复的时间而无法设置panel data，但是用iis 和tis却可以设置，这是为什么？加入了不同的地域，如果以yield为因变量，该怎么分析？请高手指点一下！

我改了一下数据。不好意思，解释一下，数据是不平衡的：一个品种可能在所有地方，所有年份都有种植，但是也可能在所有年份中只在一些地方种植，或者在所有地方却只是某些年里有种植，或者只在某些年某些地方种植。。。不知道我讲清楚了没有....哪位高手能帮我看一下原始数据？谢谢！！！

[此贴子已经被作者于2007-5-1 11:33:18编辑过]

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏0 回帖

关键词：panel data Panel model pane mode 数据 Data Panel model

you data look kind of like a three-dimension panel. but for the variety A you only have observation in the same year. so the time dimension is uselss. But you can still have a panel structure concerning variety and place. to combine year and place, you may need to refine your variables to get an easy way to deal with the data. But how to do that depends on what you want and what kind of study you are doing.

have fun.

Thank you for your advice. But is the panel structure possible concerning the variety and place but without time? What I am doing is to look for the effect of a policy on the yield during 5 years period, so a dummy variable will be included to each observation. How to combine time and place? I tried to seperate the place into dummy variables but still useless for the panel as repeated obs in each year...

First,What you want to do is a natural experiment,which occurs when some exogenous event changes the environment in which micro-individuals operate.But your data listed above seems confused, because the type of observations for A and B is not the same.Considering your purpose, A should also have 3 observations and the values of the var.year should be exactly the same as B.

Second,even if the data is OK, your purpose is not fit for a panel-data solution.Instead,it should be used in a natural experimental model.

I hope the words can help.

best

basically agree with Bu Ding. it is not neccessary to have a panel structure to do good econometric analysis. many methods employed in the panel can be done by using variations of normal cross-section econometrics. If your research is indeed a natural experiment, the most important is to justify how nature(exogenous) it is and to select a good treatment group and a good comparison group.

good luck

以下是引用布丁在2007-4-29 19:05:00的发言：

First,What you want to do is a natural experiment,which occurs when some exogenous event changes the environment in which micro-individuals operate.But your data listed above seems confused, because the type of observations for A and B is not the same.Considering your purpose, A should also have 3 observations and the values of the var.year should be exactly the same as B.

Second,even if the data is OK, your purpose is not fit for a panel-data solution.Instead,it should be used in a natural experimental model.

I hope the words can help.

best

Thank you for your comments! Actually the data is unbalanced, which means variety A appears in year1 in place1, but it may not appear in the year2 in place1, same things happened about place, due to the unavailability of data.

If using the natural experimental model, how to account for the individual and time effects, which may exist? And natural experimental model usually requires IV, what should I choose as IV? I am a beginner, could you provide me more information? Thank you!

以下是引用dgxc在2007-4-30 16:37:00的发言：

basically agree with Bu Ding. it is not neccessary to have a panel structure to do good econometric analysis. many methods employed in the panel can be done by using variations of normal cross-section econometrics. If your research is indeed a natural experiment, the most important is to justify how nature(exogenous) it is and to select a good treatment group and a good comparison group.

good luck

Thank you! How about the random coefficient model? Does it apply for repeated time-series and cross-sectional data as mine? And for natural experiment, I really do not know ho to justify how nature the data is.... Could you give me some suggestions?

In a panel data analysis, you can't use both year and place to indicate an unit, and you're oly allowed to adopt one of them. So, the words "A in year 1 and place 1" "A in year 2 and place 1" are not correct. If you want to use panel data analysis, you should say like this, "A in year 1" "A in year 2" "B in year 1" "B in year 2". Alternatively, you may apply panel data methods to other data structures, in which case you can say "A in place 1" "A in place 2" "B in place 1" "B in place 2".

Therefore, your understanding about panel data may be wrong, which leads to your intention to apply panel data analysis to your data and to your purpose, both of which don't match with panel in my opinion.

Furthermore, I think I have a clear comprehension of your purpose. Don't you want compare the control group and the treatment group to test the effects of a certain policy? If the answer is yes, then I'm sure that no other methods are better than a natural experimental model. But I have no time to explain now, and you can refer to a textbook.

[此贴子已经被作者于2007-5-1 22:27:48编辑过]

Thank you very much, Buding!

I understand what you mean about panel data, so it seems panel methods does not apply to such kind of data.

Yes, I want to test the effect of a policy, which use dummy variables. If natural experiment model helps, I will try to analyze the data with that.

I think Stata won't recognize your data as a panel. If you use some other software like Matlab, you can apply the panel methods by calculate everything explicitly.

No need to rely on panel methods if your concern is to test a policy. You have to check that the policy is really exogenous at least conditional exogenous. (that is one criterion of natural experiment). If not, you have to come up with a good IV. How to construct an IV and how to check robustness of results is a long story. It really depends on the underlying economy behind the econometrics. you may consult the following books:

1. an introductory level. upper undergraduate Introduction to econometrics by Stock and Watson. you are in China, it is easy to get a cheap copy in bookstores.

2. if you have a solid background in econometrics, you can refer to the recent book Microeconometrics: theory and application by Cameron and Trivedi. some nice guy posted this book in this forum. This is an excellent text but quite advanced.

3. if you are really good in econometrics, then go go go to James Heckman to read his papers and lecture notes.

the first book is enough for a good understanding and basic practices.

have fun