如果是非平衡的 还需要用固定效应模型和随机效应模型吗？

建议到置顶里STATA的讨论区看看那有PANEL DATA 详细说明

肯定是不一样的

for pooled and Panel:

I think u know "Panel data". But Pooled data is not a specialized definition.

for instance, u r doing a linear regression: regress consumption growth on expected income growth.

in your first estimation, you use US aggegate data.

in your 2nd effort, u could use Candian data.

on the other hand, for some reason, you may want to pool the Candian and US data together and do the regression.

to make thing clear, suppose you have 200 obs for US, and 199 obs for Canada,

then your sample size in your 1st regression is then 200, and 199 for your 2nd one.

in your 3rd, or pooled regression, the sample szie is 200+199.

now you are treating the pooled data as a sinsgle data set, and producing only one set of parameter estimations.

走过路过，千万不能错过

10# zhaosweden 受教了

同意10楼的看法

10楼很对~~再补充一下，这个例子中应该注意的是，panel data突出多个时间点对 同一群 样本的观测，而pooled data则可以是不同的样本。在10楼的例子里也得以体现，因为美国和加拿大的肯定是不同的observation~

wonderful and clearly definition

学习了

GOOD！！！！！！！！！！！

谢谢十楼的同学

有些回覆很不錯 panel的問題在於balance跟unbalance   多數工具書都只教balance  不教unbalance
你的問題應該是你的資料是unbalance 也就是在不同的期間  觀察的樣本數 是不同的  譬如在樣本期間中  企業會有新上市  或者是下市  因此樣本數會改變  這也是我一直在找的答案  我都是不去處理這些問題  因為在統計計算過程中  沒數值的會被當成na  它還是可以run  只是結果會有偏誤  但至少解決存續偏誤

英语写的简单易懂，满分！