On Mon, 31 Oct 2016, Henrique Andrade wrote:
Em 31 de outubro de 2016, Allin escreveu:
On Mon, 31 Oct 2016, Henrique Andrade wrote:
>
> Dear Gretl Team,
>>
>> I'm not an expert in MIDAS and because of that I am not confident
>> translating the following lines:
>>
>> 1. Normalized beta (zero last lag)
>> 2. spread (MIDAS)
>>
>> Line 1: Could you explain me what is "zero last lag"?
>
> The beta distribution is one parsimonious way of obtaining coefficients on
> p lags of a high-frequency regressor from either 2 or 3 hyper-parameters.
> In the 2-parameter case the coefficient on lag p (the last lag) is forced
> to zero.
The expression "zero last lag" is equivalent to "set the coefficient of
the
last lag to be equal to zero" or "coefficient of the last lag = 0"? In an
anaolgous way... The expression "non-zero last lag" is equivalent to "do
not set the coefficient of the last lag to be equal to zero" or
"coefficient of the last lag != 0"?
I'd say your alternatives amount to the same thing in context: if you
choose the version labeled "zero last lag" you're selecting a
parameterization which implies mathematically that lag p has a zero
coefficient; and if you choose "non-zero last lag" you get a
parameterization that permits a non-zero coefficient on lag p (by
shifting the whole beta distribution up or down, if the data "want"
this).
> Another line I get myself stuck is:
>>
>> 3. Decoration
>>
>> Could you give me more context?
>
> This is just used in the documentation, when categorizing matrix-related
> functions. The functions rownames() and colnames() are shown under the
> heading "Decoration".
So "decoration" here is something related to "appearance" or
"look", right?
Yes, that's right.
Allin