Purpose. The present study aimed to describe the variability of muscle thickness measures (MT) of arm and thigh muscles, obtained by ultrasound (USD) from community-dwelling healthy older women; and to obtain a predictive equation model for arm, thigh and leg MT based on upper and lower limb circumference measures using a frequentist and Bayesian approach. Methods. Sixty-two older women (74.5 ± 8.5 years) volunteered and were submitted to MT measures. Assessments of the Brachial (Br), Biceps Brachii (BB), Vastus Lateralis (VL), Rectus Femoris (RF) were obtained and grouped to infer elbow flexors (EF), and knee extensors (KE) muscles. Thigh (TC) and arm (AC) circumference were also measured. Variability of MT and circumference measures were inferred by the coefficient of variation (CV) and Huber's M-estimator. Age-adjusted linear regression models were generated to predict MT. The magnitude of evidence was obtained using Bayesian inference, determining posterior probabilities based on our data. Results. The MT measure of the EF presented lower CV (15.9%). The CV of limb circumference measures was 10.8% (TC) and 12.5% (AC) and Huber's M-estimator varied 0.18 cm (AC). Circumference measurements were significantly associated with MT with age-adjusted coefficients of determination (R 2 ) of 0.268 and 0.173 (p < 0.05), and standard error of the estimate of 0.36 cm and 0.56 cm for EF and KE, respectively. Bayesian inference confirmed the good posterior probability of the model. Conclusions. Measure of MT of the EF presented lower between-subject variability, and a better predictive equation model for MT based on AC.