Chow-Lin procedure

[Julien35]

Hi,

I'm a novice on rats and I try to interpolate a serie with the ChowLin procedure.
I have some problems with the procedure.

My serie 'Y' is bi-annual and I would like to transform it in quarterly.
I use as explicative a quarterly serie 'X'.

The code is:

@ChowLin oldser start end newser
# list of related series in regression format

then my problems are:

Does oldser serie is 'Y'? and with which frenqency?
I think this should be in quarterly data but I don't understand how could I transform Y in quarterly!

What is the "list of related series in regression format" ? just a constant and X ?

And, last question should I set FACTOR=2
Thank you for help

[moderator]

First, have you read the section of the RATS manuals on "Changing Date Frequencies" (it's Section 2.5 of the RATS Introduction if you have Version 8 )? If not, start there. As suggested there, you may want to use DISAGGREGATE (http://www.estima.com/forum/viewtopic.php?f=7&t=133) rather than CHOWLIN, as the former incorporates the features of CHOWLIN, as well as other interpolation methods.

Either way, be sure to read the comments at the top of the procedure files for additional details on using the procedure.

With respect to your questions:

"Does oldser serie is 'Y'? and with which frenqency?"

Yes, if Y is the series you want to interpolate, that would be what you use for the "oldser" parameter. With respect to the frequency, you need to read the series in at the higher frequency (e.g., quarterly) per the approach described in the manuals. The important thing to note that there is that the source data file must have dates that RATS can read and process, so that it knows you are expanding the data from the frequency indicated on the file to the (higher) frequency specified by your CALENDAR instruction. Use the VERBOSE option on DATA to verify that this is working properly.

"What is the "list of related series in regression format" ? just a constant and X ?"

From your description, yes, you'd want X as one of those series. Whether or not you want the CONSTANT depends on which model you are choosing (again, see the procedure file for details).

"And, last question should I set FACTOR=2"

Yes.

Regards,
Tom Maycock
Estima

[algrjo]

I am too am a novice at RATS and I'm trying to use the Chow-lin procudeure that I downloaded from the code section of the Estima website. However when I try to employ it to transform annual GDP into quarterly GDP using the programe below I'm getting this error message. Can you assist.

source chowlin.src
calendar 2002 1 4
all 2010: 4
open data C:\WinRATS Standard 8.1\realgdp.rats
data (for=rats,org=ROWS) / REALGDP FDIBAH PCREDITBAH RELGDPUS AIRARRIVALSBAH
@ChowLin
## OP10. Instruction INQUIRE Requires Option SERIES. If using a procedure, you may have incorrect syntax
The Error Occurred At Location 120, Line 26 of CHOWLIN

[moderator]

First, note that the ChowLin.src file is included with every copy of RATS, along with a couple hundred other procedure files. So, assuming you have a relatively recent copy of RATS, you probably already had the procedure file. The Procedure Browser on the website can be a handy way to locate procedures by function, but you may also want to refer to the PDF included with RATS that describes all the procedures shipped with the software (there are hundreds of example programs included as well).

Second, it looks like you have a space between the instruction name and the option field on your DATA instruction--if so, you need to delete that space. The option field needs to follow immediately after the instruction name.

Finally, with respect to running the procedure, anytime you are working with a procedure for the first time, you need to open the .SRC file and read the comments at the top of the procedure file. These will describe the function and proper syntax of the procedure. The syntax descriptions are organized and formatted much like what we use in the Reference Manual for built-in instructions. So, in the case of ChowLin, you'll see that the description begins:

* @ChowLin oldser start end newser
* # list of related series in regression format.

This tells you that you need to provide (at least) the original (old) series, the entry range (you can use / for the default range), and a name for the new series all on the first line, followed by a supplementary card listing the related series. You're missing the name of the original series (and the other parameters) here, which is why you get an error message.

You may also want to use one or more of the options available. Again, be sure to read the comments in the file for details.

You can also view these by adding an (ECHO) option to your SOURCE instruction:

source(echo) chowlin.src

Regards,
Tom Maycock

[algrjo]

Thank you ver much for your assistance. I have implemented your instructions. With regards to the new GDP series, should I just create a new series in the RATS data file which contains zeros and let the programme populate it with the quarterly estimates. In addition, since I am trying to convert annual GDP into quarterly GDP should I use FACTOR=1.

Thanks

[TomDoan]

Thank you ver much for your assistance. I have implemented your instructions. With regards to the new GDP series, should I just create a new series in the RATS data file which contains zeros and let the programme populate it with the quarterly estimates. In addition, since I am trying to convert annual GDP into quarterly GDP should I use FACTOR=1.

Thanks

No. That's FACTOR=4. Also, we don't recommend using the Chow-Lin interpolation procedure on a series like GDP. Chow-Lin is used in practice because procedures exist to do it, not because it's the best choice. @DISAGGREGATE with MODEL=LOGLIN and TSMODEL=RW1 should work better. The attached is a chapter on the subject from the State-Space and DSGE course (see http://www.estima.com/forum/viewtopic.php?f=24&t=528).

Interpolation and Distribution.pdf

Chapter from State-Space Course

(62.41 KiB) Downloaded 127 times