Cook's Distance in EViews

[Inns]

Hey there!
I was wondering whether it is possible to calculate Cook's Distance in EViews 7 and if so, how it could be done?
I'm writing my thesis these days so it's quite urgent.

I would really appreciate your help!

[EViews Gareth]

There isn't a built in tool for calculating it, but since Cook's distance is a simple function of studentized residuals and the hat matrix, both of which can be outputted from View->Stability Diagnostics->Influence Statistics, it should be straight forward enough to compute manually.

[Inns]

Hey Gareth!
Thanks a lot for your reply.
One of your support guys told me that there is a possibility to calculate Cook's D in EViews, but unfortunately didn't go any further...
My problem is, that there are various approaches to compute it manually, but I'm an absolute beginner. So, is it just the product of the RStudent and the Hat-Matrix? I've read somewhere, that leverage is the one to go with, i.e. RStudent x Leverage. But I'm not sure how to interpret the leverage plot or more precisely the values in brackets when clicking on them?

Do you know anything about it??

[EViews Gareth]

I'm sure your text book, or whatever reference you are using, has a description of Cook's distance.

[Inns]

As I said, there are many calculation approaches and I have no idea which one to use.

[ulrichrue]

to give you some guidance to your questions above:

Cook's Distances are defined as:
D(i)=(1/k)*(s(i)^2/s^2)*DFITS(i)^2

k...number of variables in the regression (including the constant)
s...root mean square error (=@se; S.E. of regression
in the regression output)
s(i)... root mean squared error where the ith oberservation is ommited
DFITS(i)...scaled difference in fitted values for that observation between the original
equation and an equation estimated without that observation (a series needs to be created via View/Stability Diagnostics/Influence Statistics see EVIEWS 7 User Quide II, p. 184)

s(i), the root mean squared error where the ith oberservation is ommited, is not built-in in EVIEWS as far as I know but it is included in the studentized residuals (RStudent, see equation 23.44 on p. 184 EVIEWS User Guide II), which can be created (View/Stability Diagnostics/Influence Statistics) as a series.
RStudent is defined as
RStudent(i)=e(i)/(s(i)*sqrt(1-h(i))
e(i)...these are the residuals (resid, a series which is already created)
h(i)...i-th diagonal element of the Hat Matrix (needs to be created via View/Stability Diagnostics/Influence Statistics)

rearrange equation above
s(i)=e(i)/(sqrt(1-h(i))*RSTUDENT)
to create s(i), the root mean square error, where the ith observation is ommited.

you should have created the following series now:
dffits and
s(i)
to calculate the Cook's Distances with known values for "k" and "s"

Best Regards,

Ulrich