The aim of this study was to correlate magnetic resonance spectroscopy (MRS) measurements, including that for the N-acetyl aspartate (NAA)/creatine (Cr) ratio in the vermis (denoted V-NAA), right cerebellar hemisphere (R-NAA), and left (L-NAA) cerebellar hemisphere, with the clinical scale for the assessment and rating of ataxia (SARA) score for patients with spinocerebellar ataxia (SCA) types 2, 3, and 6. A total of 24 patients with SCA2, 48 with SCA3, and 16 with SCA6 were recruited; 12 patients with SCA2, 43 with SCA3, and 8 with SCA6 underwent detailed magnetic resonance neuroimaging. Forty-four healthy, age-matched individuals without history of neurologic disease served as control subjects. V-NAA and patient age were used to calculate the predicted age at which a patient with SCA2 or SCA3 would reach an onset V-NAA value. Results showed the following: the NAA/Cr ratio decreased with increasing age in patients with SCA but not in control subjects; the SARA score increased progressively with age and duration of illness; V-NAA showed a better correlation with SARA score than R-NAA in patients with SCA2 or SCA3; the ratio of age to V-NAA correlated well with CAG repeat number; the retrospectively predicted age of onset for SCA2 and SCA3 was consistent with patient-reported age of onset; R-NAA showed a better correlation with SARA score than V-NAA in patients with SCA6; V-NAA and R-NAA correlated with clinical severity (SARA score) in patients with SCA. The correlation between CAG repeat number and age could be expressed as a simple linear function, which might explain previous observations claiming that the greater the CAG repeat number, the earlier the onset of illness and the faster the disease progression. These findings support the use of MRS values to predict age of disease onset and to retrospectively evaluate the actual age of disease onset in SCA.