Correction:

An informat like 10.7 should scale values when it processes a character string lacking an explicit decimal point, regardless of whether scientific notation is used. In Kevin's example, that happened to correspond to the non-use of scientific notation.

On Mon, 17 May 2004 16:37:16 -0400, Howard Schreier <Howard_Schreier@ITA.DOC.GOV> wrote:

>This problem got a good workover while I made my post-SUGI visit to Quebec >City. So I will comment only on the choice of informat. > >I don't think anyone suggested the simplest and most correct informat, >which is simply > > 10. > >Various people suggested 10.7, E10., and BEST10. (or minor variations). > >10.7 is simply wrong; it scales the values not expressed in scientific >notation. > >E10. works alright in this case. The documentation suggests otherwise: "The >Ew.d informat is not used extensively because the SAS informat for standard >numeric data, the w.d informat, can read numbers in scientific notation. >Use Ew.d to permit only scientific notation in your input data." So one >might expect missing values for the USA and TURKEY observations. However, >they were converted as needed. I conclude that Ew.d informats are >equivalent to their w.d counterparts. > >BESTw. is not documented as an informat, but is apparently accepted as >another synonym for W. After all, the "best" concept does not apply to >character-to-numeric conversion, and if it did apply, it would be circular >(you would need to know the numeric magnitude in order to pick the informat >used to extract the number). > >On Thu, 13 May 2004 11:42:10 -0400, CHRISTENSEN,KEVIN W <chriske2@UFL.EDU> >wrote: > >>I have a SAS dataset that looks like the following: >> >>COUNTRY POPULATION >>USA 1000000 >>CHINA 1.0016E+07 >>JAPAN 2.0616E+07 >>TURKEY 2351522 >> >>The population variable is read in as a character and I'd like to >>read it as a number. Unfortunately the scientific notation in >>some observations will make it difficult (I think). Any ideas >>around this? >>-- >>CHRISTENSEN,KEVIN W