Trim

[EViews Esther]

This thread is about the Trim add-in that performs Trimming and Winsorising.

Change log

• 2011/07/21 - Fix for a bug in winsorizing.

[basher]

Hi,
Could you suggest any concept paper(s) related to this add-in (trimming and winsorising). Thank you.
Basher

[EViews Gareth]

http://en.wikipedia.org/wiki/Winsorising :D

[ROSSINL]

Hi,

I've downloaded the trimming and winsorization add-in for Eviews 7. However, when I do winsorization on a variable, it only performs winsorization for the upper tail of the distribution. So the maximum is affected after winsorization, but the minimum is not. I've used different quantiles but the issue remains. Trimming works fine on the same data. Any suggestions how to overcome this problem?

[EViews Gareth]

Will upload a fixed version soon.

[ROSSINL]

Sure, I've attached a trimmed version of it.
I've tested it on the variables "at" and "revt" for a winsorization at 5% and 95% quintiles.
As you can see the minimum is the same for the unwinsorized and winsorized series, but the maximum differs.

PS: I see you edited your post I guess you've already located the problem? I'll wait patiently!

[EViews Gareth]

Fix is already up :D

[ROSSINL]

I can confirm that winsorizing now works flawlessly :D
Thanks for the fast response Gareth!

[dagfinnrime]

Hi,

documentation says "command line" version is "trim(q1,q2) outputname".

When I do this in a program it opens the dialog. Does "command line" literally mean the upper part of the Eviews-window, and not necessarily a program?

I see from the program that I could do (for trimming)

Code: Select all

series out = @recode(in<@quantile(in,0.05) or in>@quantile(in,0.95),na,in)

But using a command without opening the dialog box is nice.

Dagfinn

[EViews Gareth]

Open the program file and change the line that says:

Code: Select all

%outvars = @wmid(%args,2)

to be:

Code: Select all

%outvars = @wmid(%args,1)

[mhn7]

Hi,

I am currently using EViews 7 and using the trim add-on to remove outliers from my dataset. I am regression average loan balance on multiple independent variables and if I use the trim the average loan balance (from q% of highest and lowest observations at 0.99 and 0.01 respectively), then do I need to alter any of the independent variables after this?

In other words, the average loan balance will just show NA in the trimmed variables and thus I do not need to make any changes to the remaining independent variables such as cost per loan, dummies for countries etc.?

Thank you in advance!

[gerrard]

Will this add-in work with EViews 8 as well?

[trubador]

Yes. But you can always download and install the add-in to see it for yourself.

[System]

Is it possible to trim on one side? I've tried entering 0 in q% of lowest observations, but it does not allow me to do that.

[EViews Gareth]

No