Of the various kinds of doped germanium investigated for low temperature resistance thermometry, arsenic doping has been found most satisfactory in providing adequate sensitivity over the temperature range from less than 1 to a little over 100 K. Such sensors are highly reproducible as well as sensitive, but for accurate thermometry it has been a disadvantage that the relationship between sensor resistanceRand absolute temperatureTcannot be expressed accurately in analytic form. However, the relationship can be expressed within close limits over wide temperature ranges by polynomials of the formlog10R=i=0nAj(log10T)i.For an optimum representation, it is necessary to make judicious choices both of the polynomial degree and the temperature range. Insistence on a single polynomial over the entire 1–100 K range produces a solution with spurious oscillations of rms amplitude some 0.3% of the absolute temperature. Such spurious oscillations can be reduced to an amplitude of a few parts in 104of temperature by using one polynomial for the 1–20 K range and a second polynomial forT≳15 K(thus affording a small overlap region between the two polynomials). The markedly inferior results of using different temperature ranges and a different polynomial regression procedure are illustrated.