about the panel and outliers

[cocoabutter]

Hello Again!

I have two questions please
1) Suppose I run a panel, but neglect to apply and fixed/random effects across both time and cross section. What am I doing exactly here then? Am i just runing a giant/pooled OLS?

2) After estimating my panel, I was checking for outliers by examining the view, residual table. Is there an equivalent eviews equivalent function to STATA's lvr2plot command?

How do I then remove these observations? Say observation 200 for 1992. Is there an easier way then typing it in meticulously framing it? ( say in 1-10 if i want to remove observation 4 I usually type 1 3 5 10). My data is 800 observations, over 4 years
so it is structured

obs1 91
92
93
94

etc.

Thank you very much!!

[EViews Gareth]

1) Yep. As the following quick program proves:

Code: Select all

create a 1990 2010 20 series id = @obsnum series y=nrnd series x=nrnd equation panel.ls y c x pagestruct id equation giant.ls y c x show panel show giant

2) I'm not sure what LVR2PLOT does, but EViews doesn't have a way of examining outliers (yet).

An easy way of removing large outliers is to say something like:

smpl if @abs(resid)>20

[cocoabutter]

Ah ok, thank you.
So that is directly comparable with a pooled OLS, yes?

Also,do you mean sample if abs(resid)<20?

[startz]

Ah ok, thank you.
So that is directly comparable with a pooled OLS, yes?

Also,do you mean sample if abs(resid)<20?

He means "smpl"

smpl is the EViews command to set the sample.
sample is the EViews command to set the smpl.
Hey, sometimes life is counterintuitive!

[cocoabutter]

heh i was talking about the sign

[startz]

heh i was talking about the sign

Yeah, you're right about the sign.