SAS 中的回归面板数据
我现在正在查看一个面板数据集，我必须对其进行回归。由于我本学期才开始攻读博士学位并学习计量经济学课程，因此我对许多统计应用和回归方法仍然很陌生。 我想做一个简单的回归，如 Y = x1 x2 x3 等，现在我已经浏览了一些文献，发现对于面板数据，通常进行固定效应回归。另外，我的 Y 变量只有正值，所以我在考虑 Tobit 模型的方向？
我正在对金融行业分析师的报道进行一些研究。我的自变量是分析师对某家公司的覆盖范围，因此根据观察，我有 1 名分析师和 1 家公司，以及该公司的不同特征（市值和贝塔值等）。所有这些数据都是每月的。由于覆盖率不能变为负值（只能为 0），我在考虑 Tobit 模型？
您知道什么是好的回归方法吗？或者有一些好的信息来源（电子书、书面书籍，通过大学我可以访问几乎所有与我的工作领域相关的内容）（因为我确实必须学习这些东西以用于未来的研究）？
I am now looking at a panel dataset on which I have to regress. Since I only started my Phd this semester together with the econometrics courses I am still new to many statistic applications and regression methods.
I want to do a simple regression as in Y = x1 x2 x3 etc, now I already browsed through some literature and found that for panel data it's common to do a fixed effects regression. Also, my Y variable only has positive values so I was thinking in the direction of a Tobit model?
I'm doing some research concerning the coverage of analysts in the financial business. My independent variable is the coverage of analysts on a certain firm, so per observation i have 1 analyst and 1 firm, together with different characteristics(market cap and betas etc) of the firm. All this data is monthly. As coverage cannot become negative (only 0) I was thinking of a Tobit model?
Do you have any ideas what would be a good regression method? Or have some good sources (e books, written books, through university I have access to almost anything concerning my field of work) of information (cause I do have to learn these things for future research)?
如果你对这篇内容有疑问，欢迎到本站社区发帖提问 参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（1）
固定效应回归将是错误的。您的数据至少在几个月内是相关的。在 SAS/STAT 中，您将使用 proc glimmix。 SAS/ETS 可能有其他可以进行 tobit 链接的进程。也许是 proc qlim？对于一年级研究生来说，这已经相当先进了。建议您向更资深的同事寻求帮助。
Fixed effects regression will be wrong. Your data are correlated across months, at least. In SAS/STAT you would use proc glimmix. SAS/ETS may have other procs which can do tobit links. Maybe proc qlim? For a first-year grad student this is pretty advanced. Suggest you get some help from a more senior colleague.