Statistical parameters pertinent to silicon carbide fibres were estimated from experimental distributions of failure data (including strengths and strains) using the linear regression and maximum likelihood estimators. Strengths were evaluated using the average diameters of fibres, diameters at failure locations measured by SEM, and taking into account the diameter variations along the fibre axis, as evidenced by laser diffractometry. It was shown that gauge-length insensitive statistical parameters were estimated when considering that fibre fracture was dictated by two populations of partially concurrent flaw populations, including intrinsic and extrinsic flaws.