TABLE 1. RECONCILIATION OF FORTRAN PROGRAM TO SUPPLEMENTARY INFORMATION

File

Variables

Supplementary Information Columns

betas-train.out (p. 22)

one-amsetrnglob/varrawglob, *nhem, *detr, *nino, *tot

Eigenvector filerting beta

GLB, NH, DET, NIN, MULT

betas-recon1.out (p. 24)

one-amsecalglob/varrawglob, *nhem, *detr, *nino, *tot

Calibration beta:

GLB, NH, DET, NIN, MLT (cols. 7-11)

corrs-cal.out (p25)

corrsoi, corrsoi**2

 

Not reported as far as I can tell

corrs-recon.out (p. 25)

corrglob, corrglob**2, *nhem, *detr, *nino

Only the *.nino^2 is reported (column 12), but calibration period RE and r2 should be very close. I donít understand why the calibration RE [NIN] (col 12) differs so much from the calibration R2 [NIN] (col 10), and will check this more closely on another occasion.

betas-verif.out (p. 30)

one-amseverglob/varrawglob, *nhem, *nino, *tot

Verification beta:

GLB, NH, MLTA (?).

Note the absence of the verification NIN RE statistic.

corrs-verif1.out (unit=9 page 30)

corrglob, corrglob**2, *nhem,  followed by correlations for individual gridpoints.

Not reported. These are the statistics whose absence we highlighted in MM05a, MM05b.

betas-recon2.out (unit=8 page 33)

corrcal**2 (page 33)

 

Calibration g^ 2(col 13)

betas-verif2.out (unit=9 page 33)

corrcal**2 (page 34)

one-amsevertot/varrawtot (page 37)

Verification g^2 (col 19)

Perhaps calibration beta MLTB (col 17)

 

Notes: The file is from the program ftp://holocene.evsc.virginia.edu/pub/MANNETAL98/METHODS/multiproxy.f with pagination from printout in Notepad. The variables are written to the file. The Supplementary Information comes from  ftp://eclogite.geo.umass.edu/pub/mann/ONLINE-PREPRINTS/MultiProxy/stats-supp.html , with columns as shown there. At present, these identifications are based on matching of form rather than from replication of any results.