Hello everyone!
I posted the message below earlier but I have decided to change the subject and make it a bit more general and see if you guys can suggest some references.
As my message below describes, I would like to run a model in which there is a sample selection component and autocorrelation in the errors. The problem is that, because of selection, some of the past residuals are not observed.
Also, each observation is a bond issued by a given country, so observations are "clustered" both by the different countries that issue these bonds, and by the time period in which they were issued (e.g., a country may issue 3 bonds in a given quarter-year, and then none for the next two quarters, and then 2 the following quarter, and so on).
Could you suggest how I can address these problems or can you point at any references?
Thank you very much!!
Adrian
----------------------------------------
> From: [email protected]
> To: [email protected]
> Subject: st: ARCH disturbances in truncated samples?
> Date: Thu, 16 Jul 2009 17:27:49 -0400
>
> I am running a model that has the spread of a bond as a dependent variable and a bunch of domestic and global variables on the right-hand side. The spread is an inverse measure of the price for which the bond is sold to potential bondholders and gives a sense of the bond issuer's creditworthiness (higher spread indicates lower creditworthiness). Examples of domestic variables are GDP growth, debt/GDP, reserves/imports for each individual country where the bond issuer resides, while examples of global variables are the US interest rate and various liquidity and volatility indices.
>
> One problem is that I have quarterly macroeconomic data but I only observe the spread whenever countries actually issue a bond. Countries may decide not to issue a bond because the market will simply not buy the bond (for example, in the middle of a crisis) or simply because the country doesn't want to issue it (one quarter they may issue 3 bonds and the next quarter they may not need to issue at all).
>
> Now I am not exactly sure of how to model this other than with a Heckman selection model (let me know if you have other suggestions). One problem is that I think my errors are not standard Normal, iid errors... but that I think they follow some sort of ARCH structure -- for a given country, for example, the error this period may be related to the error in the previous period. The problem is that "the previous period" may not be exactly t-1... but the previous period when there was a bond issue...
>
> How do I account for this?
>
> Thank you very much in advance...
>
> Adrian
>
>
> _________________________________________________________________
> Lauren found her dream laptop. Find the PC that’s right for you.
> http://www.microsoft.com/windows/choosepc/?ocid=ftp_val_wl_290
> *
> * For searches and help try:
> * http://www.stata.com/help.cgi?search
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
_________________________________________________________________
Hotmail® has ever-growing storage! Don’t worry about storage limits.
http://windowslive.com/Tutorial/Hotmail/Storage?ocid=TXT_TAGLM_WL_HM_Tutorial_Storage_062009
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
