[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

From |
"Sergiy Radyakin" <serjradyakin@gmail.com> |

To |
statalist@hsphsun2.harvard.edu |

Subject |
Re: st: RE: Recursive program calls and the return list |

Date |
Tue, 21 Oct 2008 14:50:43 -0400 |

On Tue, Oct 21, 2008 at 1:45 PM, Nick Cox <n.j.cox@durham.ac.uk> wrote: > Notwithstanding good advice from others, it seems to me that a much > simpler way to proceed is to use globals. > ( And then you don't need -recr- to be an rclass program :) There are plenty of alternative solutions for this task: one can append rows to a matrix (matrices are 'global'), append lines in a text file, accumulate values in a list (passed from one iteration to the other), etc. In general however, recursive algorithms are not very efficient though often quite elegant. Mike might want to see if recursion is really necessary in his case. After that he will need to check if his maximum recursion depth fits into Stata's limits (which is quite moderate, IMHO): ~64: program define recurs display `1' recurs `=`1'+1' end recurs 1 2 .. 63 system limit exceeded - see manual r(1000); See: help limits I presume it is: "# of nested do-files" and in Stata 10 it is 64. If more recursion is necessary, Mike might want to switch to Mata, where the limit is much higher (thousands). Not sure how much higher, however: Stata crashes if recursion is too deep: ******************************* mata void function f(real x) { x f(x+1) } ******************************* f(1) /* crashes Stata */ which (I assume) is a bug. Until it's fixed, Mike will have to be very cautious about using recursive algorithms in Mata. Best wishes, Sergiy Radyakin > cap prog drop recr > prog recr, rclass > args count stopit > if `count' == `stopit' { > global something`count' = -99 > } > else { > global something`count' = uniform() > local k = `count' +1 > recr `k' `stopit' > } > end > // Start count at 0, run 5 times > recr 0 5 > mac list > > Nick > n.j.cox@durham.ac.uk > > Mike Lacy > > I've been experimenting with implementing recursive calls to a Stata > program, and in particular, getting the program to store something in > the return list at each call, so that at the end of program, the > return list would contain r(something1) , r(something2), ..., > r(somethingNumberOfCalls). I have experienced various difficulties > in not clobbering the return list. Here's a simple example to > demonstrate one of the problems I have had: > > /// A "do nothing" recursion example; recur calls itself "stopit" > times, > // and tries to return something each time in the locals something0, > // something1, something2, .., but only "something0" shows up in the > return list > cap prog drop recr > prog recr, rclass > args count stopit > if `count' == `stopit' { > return local something`count' = -99 > } > else { > return local something`count' = uniform() > local k = `count' +1 > recr `k' `stopit' > } > end > // Start count at 0, run 5 times > recr 0 5 > ret list > > macros: > r(something0) : ".2769154107663781" > > How could this be changed so as to accumulate something in the return > list at each recursive call? > > > * > * For searches and help try: > * http://www.stata.com/help.cgi?search > * http://www.stata.com/support/statalist/faq > * http://www.ats.ucla.edu/stat/stata/ > * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/

**References**:**st: Recursive program calls and the return list***From:*Mike Lacy <Michael.Lacy@colostate.edu>

**st: RE: Recursive program calls and the return list***From:*"Nick Cox" <n.j.cox@durham.ac.uk>

- Prev by Date:
**RE: st: RE: combining panel data values** - Next by Date:
**RE: st: RE: ivreg2 2sls, gmm2s and autocorrelation test** - Previous by thread:
**st: RE: Recursive program calls and the return list** - Next by thread:
**st: endogenous interaction term** - Index(es):

© Copyright 1996–2014 StataCorp LP | Terms of use | Privacy | Contact us | What's new | Site index |