We present experimental data and a theoretical interpretation on the conductance near the metalinsulator transition in thin ferromagnetic Gd films of thickness b ≈ 2 − 10nm. A large phase relaxation rate caused by scattering of quasiparticles off spin wave excitations renders the dephasing length L φ b in the range of sheet resistances considered, so that the effective dimension is d = 3. The observed approximate fractional temperature power law of the conductivity is ascribed to the scaling regime near the transition. The conductivity data as a function of temperature and disorder strength collapse on to two scaling curves for the metallic and insulating regimes. The best fit is obtained for a dynamical exponent z ≈ 2.5 and a correlation length critical exponent ν ′ ≈ 1.4 on the metallic side and a localization length exponent ν ≈ 0.8 on the insulating side.PACS numbers: 75.45.+j, 75.50.Cc, 75.70.Ak Since the scaling theory of Anderson localization has been proposed in 1979 [1], the metal-insulator transition in disordered conductors [2] has been one of the most extensively studied cases of quantum phase transition, both experimentally and theoretically. In its simplest form it describes non-interacting electrons in a disordered potential, where the disorder can be controlled experimentally in a variety of ways, e.g by systematic doping. One of the most dramatic predictions of the scaling theory is the absence of extended states, and therefore true metallic behavior, in systems in dimensions d ≤ 2. This has been verified in numerous experiments [3]. The other prediction is the existence of a critical point in d > 2 where the conductivity in the metallic phase goes to zero continuously with increasing disorder, in contrast to having a minimum metallic conductivity [4]. Electron-electron interactions are known to modify the behavior near a metal-insulator transition in a significant way [5]. For example, indications of a metallic state in two-dimensional systems have been found in experiment, and a number of theoretical scenarios explaining such a state have been developed [6].Near the transition, the behavior is characterized by power laws with critical exponents. For example, the dc conductivity σ(λ), with λ being a measure of disorder, follows a power law σ ∼ t s , where t = (1 − λ/λ c ) denotes the distance to the critical point at the critical disorder λ c and s is the conductivity exponent. The dynamical conductivity at the critical point, on the other hand, is characterized by the dynamical exponent z as σ(ω; λ c ) ∼ ω 1/z . The correlation length on the metallic side (λ < λ c ) diverges at the critical point as ξ ∼ t −ν ′ and the localization length (λ > λ c ) diverges as ξ ∼ |t| −ν . The critical exponents ν and ν ′ may be different. In d = 3 dimensions the relation s = ν ′ holds. The exponents in d = 3 have not been calculated in a reliable way up to now.As for any quantum phase transition, the critical exponents can not be measured experimentally at the true T = 0 critical point, but must be inferre...