Intraindividual variability and reliability in a 7-day exercise record

High levels of day-to-day or intraindividual variability implies unreliability of a measure of physical activity. Unreliability in a measure leads to attenuation of correlations with other variables. As intraindividual variability increases, the number of days necessary to assess physical activity to achieve the desired level of reliability increases. The use of an intraclass correlation to assess day-to-day reliability in a measure assumes compound symmetry.

This study reports on these issues in a sample of 165 elementary school teachers who maintained a 7-d record of physical activity each year for 3 yr. Analyses were conduced with physical activity measured as minutes, MET minutes, and kcal. Analyses were conducted with PROC MIXED in SAS controlling for the clustering effect by school.

Compound symmetry could not be supported across 7 d of the record. The weekdays tended to intercorrelate, Saturday correlated at very low levels, and Sunday correlated with Monday only. Compound symmetry was supported across the three weeks.

To achieve a reliability of 0.8 using a 7-d activity record requires 2 wk of assessment. The reliability of measures of physical activity require more careful attention, and likely require more points of assessment to achieve desired levels.