Is there a command that does forward orthogonal deviation transformation for unbalanced panel data?

[rahmat]

Hi
I am new in Stata list and have a question. I have an unbalanced panel data set and I want to do forward orthogonal deviation transformation on my data prior to use a particular estimator. I wondered if there is a command in Stata that does this for me?

[Sebastian Kripfganz]

I do not know of any command that delivers the FOD-transformed data. You may have to compute the FODs by hand. A google search delivers the following contribution to the old Statalist: http://www.stata.com/statalist/archi.../msg00283.html

[E. David Aja]

You may also be able to use -xtdpd- or user-written -xtabond2- , which both have an FOD option.

[Sebastian Kripfganz]

There exists a code by Inessa Love that does the Helmert transformation / forward-orthogonal deviations: helm.ado
You can download the program at http://www.rdecker.net/code

[Sebastian Kripfganz]

Partly inspired by this thread, I have developed a new Stata module that creates forward orthogonal deviations (and other time-series transformations):
TSTRANSFORM: new Stata module for time-series transformations of variables

[Lungile Ntsalaze]

Partly inspired by this thread, I have developed a new Stata module that creates forward orthogonal deviations (and other time-series transformations):
TSTRANSFORM: new Stata module for time-series transformations of variables

Sebastian Kripfganz

Good day sir

Thank you for developing the FOD transformation command - much appreciated. I am working on a balanced panel however when I do forward orthogonal deviation transformation, the data becomes unbalanced with 1 wave left out of the data, in addition some of the variable that contained data are now negative/missing making it impossible to run xthreg command as it works on balanced data.

How would I go about to move forward in this situation.

Thank you for your help

[Sebastian Kripfganz]

I am glad you find my tiny program useful.

Suppose you have a time / wave variable, let it be named time, and suppose for example that this time variable runs from 1 to 10. After doing all the FOD transformations that you need, you could drop the last time period / wave from your data set, because these observations will be missing for the transformed variables.

drop if time == 10

This will re-balance your data set. (Make sure to have a copy of the original data set.)

As a side comment: What is the reason of doing the FOD transformation before the use of the xthreg command? The latter applies itself a demeaning procedure to remove the fixed effects.

[dada gh]

hi; please I would like to know how can I do the helmert transformation ? Is there any command or syntax to do it ?

[Sebastian Kripfganz]

hi; please I would like to know how can I do the helmert transformation ? Is there any command or syntax to do it ?

You might find my command tstransform helpful. Among other transformations, it allows to compute the Helmert transformation (also known as forward orthogonal deviations). For more details, please see
TSTRANSFORM: new Stata module for time-series transformations of variables