Merging series - backward from an unspecified date

[maragloria]

Hi there,

I would like to merge two series, backward from an unspecified date. The first series have missing observations up to a date that can vary per country and variable. I would like to backfill these observations based on the rate of change of a 2nd series.

The difficulty for me lies in getting the observation date to start from. For instance, series 1 starts at 2004Q4 for country A, but at 2006Q1 for country B (with other dates for the remaining countries and variables). How can I get the different dates for each country, from which I should start the backfill? Could you give some tips of how best I could program this task?

many thanks!

[EViews Gareth]

Need more information on the structure of what you're trying to do. Perhaps an example workfile.

[maragloria]

Thanks for the quick reply, Gareth.

Please find a sample wf attached.

I've grouped the series I would like to merge for China and Chile as an example. It contains stock of reserves (% GDP) for a few countries, from two different sources.

For China, the series cg_rsrvsa_chn starts only in 2004Q4. I would use the observations (rate of change) from the 2nd series ns_rsrvsa_chn to backfill it.

For Chile, the starting date is 2003Q4 (cg_rsrvsa_chl). I would backfill it with ns_rsrvsa_chl.

The starting date can vary for the remaining countries.

[EViews Gareth]

So assuming you have some sort of loop that goes through the countries, you can use the series.@first data member to retrieve the date of the first non-NA observation and then use that to set the sample:

Code: Select all

%start = cg_rsrvsa_{%country}.@first
smpl @first {%start}


[maragloria]

Quite simple! Thank you for the suggestion. I've implemented it in a for loop for the remaining countries.

The backward filling (staring from later date and moving back up to first date) proved a bit more challenging. I've implemented with a while loop:

Code: Select all

   smpl @all
   series cg_{%country}_{%pays} =  {%1}_{%country}

   'get date of the first non-NA observation
   %start = cg_{%seriesname}_{%country}.@first   
   smpl @first {%start}
   if @obssmpl > 1 then
      !temp = @obssmpl
      while !temp > 1
         smpl @first @first+!temp-2   
         series cg_{%seriesname}_{%country} = cg_{%seriesname}_{%country}(+1)*({%2}_{%country}/{%2}_{%country}(+1)) 'backward filling
         !temp=!temp-1
      wend
   endif

Thanks once again!