A detailed phenomenological re-analysis of previously published conductivity data, σ(T, x), is presented. It was shown in [1] that the cusp-like low-temperature contribution can be described by $$\bar \beta (x) \cdot T^p $$ wherep=0.19±0.03. Starting from this result, two furtherT dependent contributions are separated: The high-temperature region is dominated by a positive contribution σht (T, x), which is approximately independent ofx, nearly linear inT above 100 K and nearly quadratic inT below 30 K. ForT≲ 4 K, there is a small deviation, increasing withT, from the superposition of the above mechanisms. The relation between $$\bar \beta $$ , being negative, and theT independent part, $$\bar \alpha $$ , exhibits a singularity, $$\bar \beta \sim (\bar \alpha - \bar \alpha _0 )^\vartheta $$ where $$\bar \alpha _0 = 270 \pm 50 \Omega ^{ - 1} cm^{ - 1} $$ andϑ=0.68±0.05 −(p−0.19). This singularity should be related to the metal-semiconductor transition, taking place atT c ≈0.14. The quantity $$\bar \alpha _0 $$ should be interpreted as minimum metallic conductivity. The limitingT dependences asx→x c +0 agrees quantitatively with that one obtained previously for the activated region,x→x c −0. Extrapolation of the phenomenological model obtained leads to the hypothesis that the interplay of $$\bar \beta \cdot T^p $$ and σht could be the main origin of the temperature coefficient changing its sign in the Mooijregion, at σ tc=0. The model enables several trend predictions concerning the value of σ tc=0.