#!/bin/sh ----cut here-----cut here-----cut here-----cut here----- # This is a shell archive. # Run the following text with /bin/sh to extract. YASASTART=`pwd` echo yasa: extracting info.numbers.doc cat - << \Funky!Stuff! > info.numbers.doc Non-zero Error Codes from dptpss Code Occured In Meaning -2 dvmin log(nobs*lamhat) > lamlim(2) (not fatal) - -1 dvmin log(nobs*lamhat) < lamlim(1) (not fatal), will occur - if limits on lambda hat are input by user and are equal. 1 dptpss dimension error (nobs < 0 or m < 0 or dim < 0 - - - or ntbl < 0 or ntbl > ldtbl or 2*m - dim < 0 ) - 2 dptpss error in dreps, the first ncov1 columns of s do not duplicate the replication structure of des 3 dptpss lwa (length of work) < ncts1*(nobs + nuobs + 1) + npar*(nobs + npar + 1) + (npar - nnull)* (npar- 2*nnull+2+nobs)+npar+nobs 4 dptpss liwa (length of iwork) < 3*nobs - ncts1 5 dptpss error in dmaket (error in creation of T) 6 dptpss sigma is rank deficient 1001* dsnsm dimension error (nobs < 0 or npar < 0 or nnull < - - - 0 or (npar - nnull) < 0 ) - 1002* dsnsm lwa (length of work) < (npar - nnull)*(npar - 2*nnull + 2 + nobs) + npar + nobs 1003* dsnsm liwa (length of iwork) < 2*npar - nnull 1004* dsnsm error in ntbl (ntbl < 0 or ntbl > ldtbl) 1101* ddcom dimension error (nobs < 0 or npar < 0 or nnull < - - - 0 or (npar - nnull) < 0 or ldx < nobs or ldx < - npar) 2 1102* ddcom ldcaux (length of dcaux) < (npar - nnull) + 2*npar - nnull 1103* ddcom lwa (length of work) < (npar - nnull)* (nobs - nnull + 1) + nobs 1104* ddcom liwa (length of iwork) < npar - nnull 1111* dsgdc calculated nnull is smaller than input nnull 1121 dcrtz error in dtrsl, R is singular 1131* dzdc lwa (length of work) < (npar - nnull)* (nobs - nnull) + npar - nnull + nobs 1132 dzdc transpose of J is necessary (npar > nobs) but npar 2 > ldx 1133* dzdc failure to converge in dsvdc called from dtsvdc (us- ing J ') 2 1134 dzdc failure to converge in dsvdc (using J '), this error 2 can usually be cured by increasing the parameter MAXIT in the LINPACK routine dsvdc 1135* dzdc failure to converge in dsvdc called from dtsvdc 1136 dzdc failure to converge in dsvdc, this error can usually be cured by increasing the parameter MAXIT in the LINPACK routine dsvdc 1201* dgcv dimension error ( nobs < 0 or npar < 0 or nnull < - - - 0 or (npar - nnull) < 0 ) - 1202* dgcv error in ntbl (ntbl < 0 or ntbl > ldtbl) 2 1203* dgcv ldcaux (length of dcaux) < (npar - nnull) + 2*npar - nnull 1204* dgcv lwa (length of work) < (npar - nnull) + nobs 1205 dgcv lamlim(1) > lamlim(2) 1211 dvlop svals(1) = 0.0d0 1212* dvlop npsing is incorrect 1213* dvlop lamlim(1) > lamlim(2) 1221 dcfcr error in dtrco, G is singular 1222 dcfcr error in dtrsl, R is singular *Should not occur in normal circumstances Non-zero Error Codes from dsnsm Code Occured In Meaning -3 dsgdc calculated nnull is larger than input nnull (not fatal) -2 dvmin log(nobs*lamhat) > lamlim(2) (not fatal) - -1 dvmin log(nobs*lamhat) < lamlim(1) (not fatal), will occur - if limits on lambda hat are input by user and are equal. 1 dsnsm dimension error (nobs < 0 or npar < 0 or nnull < - - - 0 or (npar - nnull) < 0 ) - 2 dsnsm lwa (length of work) < (npar - nnull)* (npar- 2*nnull+2+nobs)+npar +nobs 3 dsnsm liwa (length of iwork) < 2*npar - nnull 4 dsnsm error in ntbl (ntbl < 0 or ntbl > ldtbl) or tau <0 101 ddcom tau <0 or dimension error (nobs < 0 or npar < 0 - - or nnull < 0 or (npar - nnull) < 0 or ldx < nobs - - or ldx < npar) 2 102* ddcom ldcaux (length of dcaux) < (npar - nnull) + 2*npar - nnull 103* ddcom lwa (length of work) < (npar - nnull)* (nobs - nnull + 1) + nobs 104* ddcom liwa (length of iwork) < npar - nnull 111 dsgdc calculated nnull is smaller than input nnull 121 dcrtz error in dtrsl, R is singular 131* dzdc lwa (length of work) < (npar - nnull)* (nobs - nnull) + npar - nnull + nobs 132 dzdc tau < 0 133 dzdc transpose of J is necessary (npar > nobs) but npar 2 > ldx 134 dzdc failure to converge in dsvdc called from dtsvdc (us- ing J '), this error can usually be cured by increas- 2 ing the parameter MAXIT in the LINPACK routine dsvdc or by increasing the value of the argument tau to dsnsm (see discussion in TR775 (rev.)) 135 dzdc failure to converge in dsvdc (using J '), this error 2 can usually be cured by using the truncation option or by increasing the parameter MAXIT in the LINPACK routine dsvdc 136 dzdc failure to converge in dsvdc called from dtsvdc, this error can usually be cured by increasing the parameter MAXIT in the LINPACK routine dsvdc or by increasing the value of the argument tau to dsnsm (see discussion in TR775 (rev.)) 137 dzdc failure to converge in dsvdc (using J '), this error 2 can usually be cured by using the truncation option or by increasing the parameter MAXIT in the LINPACK routine dsvdc 201* dgcv dimension error (nobs < 0 or npar < 0 or nnull < - - - 0 or (npar - nnull) < 0 ) - 202* dgcv error in ntbl (ntbl < 0 or ntbl > ldtbl) 2 203* dgcv ldcaux (length of dcaux) < (npar - nnull) + 2*npar - nnull 204* dgcv lwa (length of work) < (npar - nnull) + nobs 205 dgcv lamlim(1) > lamlim(2) 211 dvlop svals(1) = 0.0d0 212 dvlop npsing is incorrect 213* dvlop lamlim(1) > lamlim(2) 221 dcfcr error in dtrco, G is singular 222 dcfcr error in dtrsl, R is singular *Should not occur in normal circumstances Non-zero Error Codes from dtpss Code Occured In Meaning -2 dvmin log(nobs*lamhat) > lamlim(2) (not fatal) - -1 dvmin log(nobs*lamhat) < lamlim(1) (not fatal), will occur - if limits on lambda hat are input by user and are equal. 1 dtpss dimension error (nobs < 0 or m < 0 or dim < 0 - - - or ntbl < 0 or ntbl > ldtbl or 2*m - dim < 0 ) - 2 dtpss error in dreps, covariates do not duplicate the re- plication structure of des 3 dtpss lwa (length of work) < nuobs*(2 + ncts + nuobs) + nobs 4 dtpss liwa (length of iwork) < 2*nobs + nuobs - ncts 11 dsetup error in dmaket (error in creation of T) 101* dsgdc1 lwa < 2*npar 102 dsgdc1 F 'KF is not of full rank 2 2 103 dsgdc1 failure to converge in dsvdc, this error can usually be cured by increasing the parameter MAXIT in the LINPACK routine dsvdc 201* dgcv1 dimension error ( nuobs < 0 or ncts1 < 0 or nuobs - - - ncts1 < 0 ) - 202* dgcv1 error in ntbl ( ntbl < 0 or ntbl > ldtbl ) 203* dgcv1 lwa (length of work) < nuobs - ncts1 + nobs 204 dgcv1 lamlim(1) > lamlim(2) 211 dvlop svals(1) = 0.0d0 212* dvlop npsing is incorrect 213 dvlop lamlim(1) > lamlim(2) 221 dcfcr1 error in dtrco, G is singular *Should not occur in normal circumstances Non-zero Error Codes from dpred Code Occured In Meaning 1 dpred dimension error ( nobs < 0 or nct < 0 or m < 0 or - - - dim < 0 ) - 2 dpred npar != ndesb + nct + ncov1 + ncov2 3 dpred lwa (length of work) < npred * ( nct + ndesb ) 4 dpred error in dmaket (error in creation of T) Funky!Stuff! echo yasa: extracting major.changes.doc cat - << \Funky!Stuff! > major.changes.doc This is a listing of the major changes to Gcvpack for the second release. Users who call only the drivers dptpss.f and dtpss.f will not have to make any modifications to their calling routines. Users who call the driver dsnsm.f must add the argument factor to the calling sequence (start with a value of 1.0) and must declare dout to be of length 5 rather than 4. All users must note that the search for lambda hat is now done in the log base 10 scale rather than the natural log scale. This will change the output from all drivers. Users who call any other routines must be very careful to note the changes in calling sequences listed below. We do not expect to release another version of Gcvpack after this one. You must replace all of the release 1 routines with release 2 routines to insure correct computation. The set of necessary linpack routines has not changed. affected routines description ddsdc.f+, ddcom.f, ddsdc.f is replaced with dcrtz.f, dzdc.f. dcrtz.f**, dzdc.f** ddcom.f, dzdc.f Nonzero info and svals from dsvdc are now returned in npsing and svals. dsgdc1.f Nonzero info and svals from dsvdc are now returned in p and svals. dptpss.f, dsnsm.f, dtpss.f Nonzero info and svals from dsvdc are now returned in iout(1) and svals. dpdcr.f*, ddiag.f**, Calculation of the diagonal of the hat ma- dgcv.f, dgcv1.f trix has been removed from dpdcr.f and is now done in the new routine ddiag.f. dtpss.f, dsnsm.f, dptpss.f, Iout and dout are now assigned as early as dgcv1.f, dgcv.f possible to return more information when info is positive. dgcv.f, dgcv1.f, dpmse.f, Lambda values are expressed as log base 10 dptpss.f, dsnsm.f, dtpss.f, rather than natural log. dvl.f, dvlop.f dsuy.f*, dreps.f*, dtpss.f, The argument dfrep is removed from the dptpss.f calling sequences of dsuy.f and dreps.f. dvlop.f*, dgcv1.f, dgcv.f The argument npar is removed from the cal- ling sequence of dvlop.f. dcfcr.f*, dgcv.f The argument fgaux is removed from the calling sequence of dcfcr.f. ddcom.f, dcrtz.f* The argument lwa is removed from the cal- ling sequence of dcrtz.f. ddcom.f*, dsnsm.f*, dzdc.f* The argument factor is added to allow con- trol over the amount of truncation. dsnsm.f The argument dout must be declared with length 5 rather than length 4. The trun- cation ratio is now returned in dout(5). ddcom.f*, dzdc.f*, The frobenius norm of R - R is now re- k dtsvdc.f* turned as an argument. dcfcr.f, dcfcr1.f, The routine dgemv.f is called to perform dgemv.f**, ddiag.f, matrix multiplication. (Dgemv.f is from dpdcr.f, dpmse.f, dpred.f, the extended BLAS). drsap.f testtpss.out, testptpss.out Changes due to the switch from natural log to log base 10 and reformatting of output. fort.8+, testtpss.in**, The input files testtpss.in and testptpss.in testptpss.in**, testtpss.f, replace fort.8 and are now read as standard testptpss.f input. inteqn.f**, inteqn.in**, Routines and files used for new integral inteqn.out**, mktpar.f**, equation test driver inteqn.f. mkxys.f** * Calling sequence changed ** New routine or file + Deleted routine or file Funky!Stuff!