2 The [Fe/H] catalog

2.1 Contents

The [Fe/H] catalog contains entries for 1117 G and K stars with luminosity classes from II through IV. The contents of the catalog include

1.

mean values of [Fe/H],

2.

rms errors for those means,

3.

numbers of degrees of freedom,

4.

numbered comments,

5.

literature references, and

6.

a FORTRAN search program.

Reddenings derived for particular stars are included in the numbered comments. The literature references include remarks about the treatment of the input data for the catalog. The references include a number of potential data sources for the catalog that were not used. Reasons for not using those data are given with the references. Catalog users who wish to know why a particular reference was not used may find the reference by searching for the name of its first author.

The search program deserves special attention. Given an input list of HD or comparable numbers, the program compiles three output lists. One list contains primary data from the catalog, while a second list contains numbered comments. The third list includes all literature sources from which the catalog values of [Fe/H] for the input list of stars have been drawn. The search program is included for two reasons: to make it possible to use the catalog conveniently, and to encourage catalog users to cite original sources for the catalog data.

Users with questions about the analysis used to produce the catalog are encouraged to consult the following sources: Taylor (1998a) for the preparation of the analysis, Taylor (1998b) for values of [Fe/H] for standard stars, and Taylor (1999) for a description of the analysis itself and its basic results.

2.2 Statistics

Some users may find the number of degrees of freedom to be unfamiliar. To show one way in which those data may be applied, the important problem of comparing catalog values of [Fe/H] may be considered. Given entries for stars 1 and 2, one calculates

$$\Delta\mathrm{[Fe/H]} = \vert\mathrm{[Fe/H]}_1 - \mathrm{[Fe/H]}_2\vert,$$ (1)

$$\sigma^2=\sigma_1^2 + \sigma_2^2,$$ (2)

$$t=\Delta\mathrm{[Fe/H]}/\sigma,$$ (3)

and

$$\nu = (\sigma_1^2+\sigma_2^2)^2 \big/ (\sigma_1^4/\nu_1+\sigma_2^4/\nu_2).$$ (4)

In these equations, $\sigma$ is rms error, $\nu$ is number of degrees of freedom, and quantities without subscripts are net quantities. When values of t and $\nu$ have been obtained, they are used as arguments in a "two-tailed'' t table. If [Fe/H]₁ and [Fe/H]₂ differ significantly, the confidence level C obtained from the table should equal or exceed C₀, where

C₀^N = 0.95

and N is the number of values of $\Delta\mathrm{[Fe/H]}$ that are currently being tested. (For more about t tests, see pp. 197-201 of Bethea et al. 1985. For a derivation of Eq. (5), see Appendix A of Taylor 1996).