Long and short of nerve conduction measures: reproducibility for sequential assessments