DOFOR LABEL

[comac]

Hi there, I am trying to use DOFOR with labels or variables' names.
In the simplest case, I have two series named deUK and deJP which I intend to use as dependent variables. My regressors are named bUK and bJP.
Since 'UK' and 'JP' are recurring in my variables' names, I would like to implement something doing the following. All in one step:

Code: Select all

linreg deUK
#constant bUK

AND

linreg deJP
#constant bJP


I have tried something like this, but obviously this is not the right way in RATS:

Code: Select all

declare vector[labels] vl(2)
input vl
"UK" "JP"

dofor i = 1 2
linreg de'vl(i)'
#constant bvl(i)
end do for i


Many thanks in advance!

[TomDoan]

Use the %S function to create the required series names:

Code: Select all

dofor [label] co = "uk" "jp"
   linreg %s("de"+co)
   # constant %s("b"+co)
end dofor

[economics2012]

Hi,

I need to do a dofor loop for different series: aggregate, Agric, Food and Soda.
i.e. I need to repeat the linear regression below for Agric, Food and Soda,

Code: Select all

open data "aggregate.txt"
calendar(m) 1973:1
data(format=prn,nolabels,org=columns) 1973:01 2009:12 date oil o1 o2 o3 aggregate Agric   Food Soda

smpl 1973:1 2009:12

 ieval nlag=12

display 'use specification, nw errors'
linreg(robusterrors,lags=13,lwindow=neweywest) aggregate
# constant aggregate{1 to nlag} oil{1 to nlag} o1{1 to nlag}
exclude
# o1{1 to nlag}

display ' specification for 2-period, nw errors'
linreg(robusterrors,lags=13,lwindow=neweywest) aggregate
# constant aggregate{2 to nlag+1} oil{2 to nlag+1} o1{2 to nlag+1}
exclude
# o1{2 to nlag+1}

Any help?

[TomDoan]

Hi,

I need to do a dofor loop for different series: aggregate, Agric, Food and Soda.
i.e. I need to repeat the linear regression below for Agric, Food and Soda,

Code: Select all

open data "aggregate.txt"
calendar(m) 1973:1
data(format=prn,nolabels,org=columns) 1973:01 2009:12 date oil o1 o2 o3 aggregate Agric   Food Soda

smpl 1973:1 2009:12

 ieval nlag=12

display 'use specification, nw errors'
linreg(robusterrors,lags=13,lwindow=neweywest) aggregate
# constant aggregate{1 to nlag} oil{1 to nlag} o1{1 to nlag}
exclude
# o1{1 to nlag}

display ' specification for 2-period, nw errors'
linreg(robusterrors,lags=13,lwindow=neweywest) aggregate
# constant aggregate{2 to nlag+1} oil{2 to nlag+1} o1{2 to nlag+1}
exclude
# o1{2 to nlag+1}

Any help?

Code: Select all

dofor s = aggregate Agric Food Soda
   linreg(robusterrors,lags=13,lwindow=neweywest) s
   # constant s{1 to nlag} oil{1 to nlag} o1{1 to nlag}
   exclude
   # o1{1 to nlag}
   display ' specification for 2-period, nw errors'
   linreg(robusterrors,lags=13,lwindow=neweywest) s
   # constant s{2 to nlag+1} oil{2 to nlag+1} o1{2 to nlag+1}
   exclude
   # o1{2 to nlag+1}
end door


Why are you doing Newey-West with 13 lags on the first regression? That should have serially uncorrelated errors.

[economics2012]

Yes you are absolutely correct. For h>1, the errors are serially correlated.

[economics2012]

One more question,
When I run the code for 50 different sectors,

Code: Select all

open data "dataset.txt"
calendar(m) 1973:1
data(format=prn,nolabels,org=columns) 1973:01 2009:12 date oil o1 o2 o3
aggregate agriculture Food Soda Beer Smoke Toys Fun Books Hshld Clths
Hlth MedEq Drugs Chems Rubbr Txtls BldMt Cnstr Steel FabPr Mach ElcEq
Autos Aero Ships Guns Gold Mines Coal Oil Util Telcm PerSv BusSv Hardw
Softw Chips LabEq Paper Boxes Trans Whlsl Rtail Meals Banks Insur RlEst
Fin Other

smpl 1973:1 2009:12

dofor s = aggregate agriculture Food Soda Beer Smoke Toys Fun Books Hshld Clths
Hlth MedEq Drugs Chems Rubbr Txtls BldMt Cnstr Steel FabPr Mach ElcEq
Autos Aero Ships Guns Gold Mines Coal Oil Util Telcm PerSv BusSv Hardw
Softw Chips LabEq Paper Boxes Trans Whlsl Rtail Meals Banks Insur RlEst
Fin Other




I got this error:

## I1. Expected Instruction - AGG Is Not Recognizable As One
>>>>aggregate <<<<

I think the line is too long or so...

I truly appreciate your help.

[TomDoan]

Some programming languages use a ; to indicate that a line is done, and keep running a single logical line until they see it. RATS uses the alternative of using a symbol ($ in our case) as the end of a line that isn't done. So you want

dofor s = aggregate agriculture Food Soda Beer Smoke Toys Fun Books Hshld Clths $
Hlth MedEq Drugs Chems Rubbr Txtls BldMt Cnstr Steel FabPr Mach ElcEq $
Autos Aero Ships Guns Gold Mines Coal Oil Util Telcm PerSv BusSv Hardw $
Softw Chips LabEq Paper Boxes Trans Whlsl Rtail Meals Banks Insur RlEst $
Fin Other

with something similar on the DATA instruction.

[economics2012]

Is there a way I can get the p-values results for 50 sectors across 12-horizons in one table?

It is a pain to copy each one of them into excel.

Thanks a lot.

[TomDoan]

Code: Select all

report(action=define)
dofor s = aggregate agriculture Food Soda Beer Smoke Toys Fun Books Hshld Clths $
Hlth MedEq Drugs Chems Rubbr Txtls BldMt Cnstr Steel FabPr Mach ElcEq $
Autos Aero Ships Guns Gold Mines Coal Oil Util Telcm PerSv BusSv Hardw $
Softw Chips LabEq Paper Boxes Trans Whlsl Rtail Meals Banks Insur RlEst $
Fin Other
   report(row=new,atcol=1) %l(s)
   do h=0,11
       do the calculation for that horizon
       report(row=current,atcol=h+2) whatever_the_pvalue_is_for_h
   end do h
end do s
report(action=show,format=xls)

This will prompt you for the name of the Excel file you want to create. You could also put in an

open copy mytable.xls

command and use UNIT=COPY, or you could just do ACTION=SHOW without the FORMAT=XLS, and it will put the table up on the screen. You can then copy and paste the whole thing into Excel.

[economics2012]

Hi Tom,


What should I put for: whatever_the_pvalue_is_for_h?

I really appreciate your help

[TomDoan]

The idea was to avoid the copy-paste-edit steps for all the horizons. For a given value of h, you need

linreg(robusterrors,lags=13,lwindow=neweywest) s
# constant s{1+h to nlag+h} oil{1+h to nlag+h} size{1+h to nlag+h}
exclude
# size{1+h to nlag+h}

And the whatever_the_pvalue_is_for_h is just %SIGNIF

[economics2012]

Perfect. Thanks a lot, I truly appreciate it.

But in this case, I am doing Newey-West with 13 lags on the first regression which has serially uncorrelated errors. So should I change it?

[TomDoan]

Change to

linreg(robusterrors,lags=%if(h==0,0,13),lwindow=neweywest) s

which will give you "White" standard errors for h=0 and NW for h>0.

[economics2012]

Perfect. Thank you so much. You are very helpful.