[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

From |
"Terra Curtis" <terra.curtis@cambridgefinance.com> |

To |
<statalist@hsphsun2.harvard.edu> |

Subject |
RE: st: RE: RE: Bracket [ ] notation |

Date |
Wed, 10 Aug 2005 16:43:35 -0400 |

Ok. Thanks for your help. -----Original Message----- From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Nick Cox Sent: Wednesday, August 10, 2005 4:40 PM To: statalist@hsphsun2.harvard.edu Subject: RE: st: RE: RE: Bracket [ ] notation I can only repeat advice from previous emails. 1. -capture- was what you asked for, but I indicated that it wasn't likely to be what you really need. 2. Unbalanced panels are not exotic, and Stata should be able to handle your problems without your needing to learn or apply any special tricks, but this should be much easier if you use time-series operators. That would make the -by:- structure unnecessary. I'm afraid Statalist isn't very efficient in cases like yours whenever much more detailed advice requires us to know the data and the model in the way that you do. (Nor am I asking for that.) Nick n.j.cox@durham.ac.uk Terra Curtis > -capture- didn't seem to help. > > I think what's happening is, since I have unbalanced panel data, the > nonlinear equation I have specified is trying to include observations > for a company that doesn't have an observation in that year, even > though some of the companies will have an observation in that (or any > given) year. > > I've tried putting an if statement at the end of the -by > coname: replace- > line saying if *none of the variables in the specified equation are > blank.* But this doesn't seem to make a difference. > > Is the problem more clear now? And if so, do you know how I could deal > with it? n j cox > -capture- gives you a way to ignore a command if it produces an error > and would otherwise stop your program. I have to guess that is not > what you need here, but I'm less clear about what you do need. > > >>> Terra Curtis > > Is there any way to tell Stata to ignore the command when at least one > of the referenced entries does not exist? > > Nick Cox > > If the subscript on the RHS is missing (meaning not present in the > dataset), the result on the LHS will be missing too (meaning > "missing"). > > Here the use of -by:- is just a detail and does not affect the > principle. > The principle can be seen by > > sysuse auto > di mpg[0] > di mpg[75] > > As advised in another thread, you would surely be better off > re-casting this using time-series operators. > > Nick > n.j.cox@durham.ac.uk > > Terra Curtis > > > I am running into some problems trying to run a nonlinear > regression. > > I continue to get error 480 "starting values invalid or some RHS > > variables have missing values." I think I see where my error is but > > I'm not sure how STATA deals with a reference, say 'by varname: > > data[x]' if x is out of the range of data for that varname. Here > is > my code: I have written a program to specify the nonlinear > equation, > and that is where the error is: > > > > program nlstk2_depreciation > > version 8.1 > > if "`1'" == "?" { > > global S_1 "B3 B4 B5 drd2 dmktg2" > > global B3=1 > > global B4=1 > > global B5=1 > > global drd2=0.15 > > global dmktg2=0.15 > > exit > > } > > forvalues x=10/53 { > > by coname: replace `1'=$B3+$B4*(ireal46[`x'] + > > (1-$drd2)*ireal46[`x'-1] + (1-$drd2)^2*ireal46[`x'-2] + > > (1-$drd2)^3*ireal46[`x'-3] + (1-$drd2)^4*ireal46[`x'-4] + > > (1-$drd2)^5*ireal46[`x'-5] + (1-$drd2)^6*ireal46[`x'-6] + > > (1-$drd2)^7*ireal46[`x'-7]+ (1-$drd2)^8*ireal46[`x'-8] + > > (1-$drd2)^9*ireal46[`x'-9]/$drd2) + $B5*(imktg[`x'] + > > (1-$dmktg2)*imktg[`x'-1] + (1-$dmktg2)^2*imktg[`x'-2] + > > (1-$dmktg2)^3*imktg[`x'-3] + (1-$dmktg2)^4*imktg[`x'-4] + > > (1-$dmktg2)^5*imktg[`x'-5] + (1-$dmktg2)^6*imktg[`x'-6] + > > (1-$dmktg2)^7*imktg[`x'-7] + (1-$dmktg2)^8*imktg[`x'-8] + > > (1-$dmktg2)^9*imktg[`x'-9]/$dmktg2) if `x'==year > > } > > end > > > > I have panel data for companies over a period of years, although > each > company has a different amount of years of data (no company > has more > than > 53 years of data). If a company has 10 years of data, I want the > > regression to use all 10 years of data. If it has more than 10 years > > of data, I want it to use the most recent 10 (say, years 1990-1999), > > then use the next 10 (years 1989-1998), and so on until it gets to > the > last set of 10 years for that company. I think the logic of > the > > equation I > have written works out, so I'm guessing the error is coming > > from some > referencing problem in the brackets...I don't quite > understand how > STATA would deal with ireal46[49] if some company > didn't have a > 49th ireal46 entry. > > * > * For searches and help try: > * http://www.stata.com/support/faqs/res/findit.html > * http://www.stata.com/support/statalist/faq > * http://www.ats.ucla.edu/stat/stata/ > > > * > * For searches and help try: > * http://www.stata.com/support/faqs/res/findit.html > * http://www.stata.com/support/statalist/faq > * http://www.ats.ucla.edu/stat/stata/ > * * For searches and help try: * http://www.stata.com/support/faqs/res/findit.html * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/ * * For searches and help try: * http://www.stata.com/support/faqs/res/findit.html * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/

**References**:**RE: st: RE: RE: Bracket [ ] notation***From:*"Nick Cox" <n.j.cox@durham.ac.uk>

- Prev by Date:
**RE: st: RE: RE: Bracket [ ] notation** - Next by Date:
**st: RE: RE: time dummy or trend** - Previous by thread:
**RE: st: RE: RE: Bracket [ ] notation** - Next by thread:
**st: RE: time dummy or trend** - Index(es):

© Copyright 1996–2016 StataCorp LP | Terms of use | Privacy | Contact us | What's new | Site index |