Generate Stata date variables from run-together date variables --------------------------------------------------------------
^todate7^ varlist [^if^ exp] [^in^ range], ^g^enerate^(^newvarlist^)^ ^p^attern^(^patternspec^)^ [ ^f^ormat^(^format^)^ ^c^end^(^year^)^ ]
Description -----------
^todate7^ takes run-together date variables consisting of integers or integer characters and generates the corresponding Stata date variables.
Note that ^todate7^ was written for Stata 7. As of September 2005, users of Stata 8 or later should switch to using ^todate^.
Remarks -------
As explained in [U] 27.2.2.2, Stata's ^date()^ function cannot convert run-together date variables which consist of integers or integer characters and represent some permutation of years, months and days, with values such as ^19520121^ or ^520121^ or ^"19520121"^, to Stata date variables. However, it is possible to use other numeric or string functions to extract year, month and day and then to compute the dates.
A similar problem and a similar solution arise with run-together dates representing years and one of half-years, quarters, months or weeks.
^todate7^ offers an alternative. It can handle one or more numeric and/or string variables with run-together dates which are integers or integer characters, hereafter both referred to as digits.
Options -------
^generate(^newvarlist^)^ specifies a list of new variable names to hold the Stata date variables. There must be as many new names as there are existing run-together date variables. This is a required option.
^pattern(^patternspec^)^ specifies the pattern for interpreting digits in the dates. ^y^ indicates ^y^ear. ^h^ indicates ^h^alf-year. ^q^ indicates ^q^uarter. ^m^ indicates ^m^onth. ^w^ indicates ^w^eek. ^d^ indicates ^d^ay.
The patterns supported are of types ^mdy^, ^yh^, ^yq^, ^ym^ and ^yw^. For example, type ^mdy^ consists of some permutation of ^m^s, ^d^s and ^y^s. In general, all ^y^s, ... , all ^d^s should be contiguous. Thus ^p(yyyymmdd)^ indicates that ^19520121^ is to be interpreted as 1952 January 21. Dates like ^3281952^ and ^12251999^ for March 28 1952 and December 25 1999, in which the first digit for shorter dates is not specified, but could be ^0^, should be specified as ^p(mmddyyyy)^. ^todate7^ cannot handle input variables whose maximum length does not match the length of pattern specified, nor input variables whose non- missing values vary in length by more than 1 digit. Such variables are skipped. This is a required option.
^cend(^year^)^ specifies the end of the century (any period of 100 years) to which years belong. With ^cend(2000)^ year digits of ^52^ and ^99^ will be interpreted as ^1952^ and ^1999^. With ^cend(2060)^ they will be interprete > d as ^2052^ and ^1999^. ^cend()^ is ignored if 4 digit years are given.
^format(^format^)^ specifies a format to be attached to each new variable, usually but not necessarily a date format. See help on @dfmt@ or @tfmt@.
Examples --------
. ^todate7 d1 d2 d3, gen(ndate1-ndate3) p(yyyymmdd) f(%dd_m_cy)^ . ^todate7 d1 d2 d3, gen(ndate1-ndate3) p(mmddyy) c(2000) f(%dd_m_cy)^ . ^todate7 d1 d2 d3, gen(ndate1-ndate3) p(yyyyq) f(%tq_q_cy)^
Author ------
Nicholas J. Cox, University of Durham, U.K. n.j.cox@@durham.ac.uk
Acknowledgments ---------------
Kit Baum, Roger Harbord and Gary Longton made very helpful comments.
Also see --------
On-line: help for @datefcn@, @ywfcns@, @dfmt@, @tfmt@ Manual: ^[U] 27 Commands for dealing with dates^ FAQ: @http://www.stata.com/support/faqs/data/dateseq.html@