Multivariate examination of data from gait analysis of persons with stroke

Phys Ther. 1998 Aug;78(8):814-28. doi: 10.1093/ptj/78.8.814.

Abstract

Background and purpose: Gait analyses yield redundant information that often is difficult to interpret. The purpose of this study was to show how principal-component analysis can provide insight into gait data obtained from persons with stroke.

Subjects: Twenty male and 11 female adults who were ambulatory were studied (mean age = 60.5 years, SD = 11.8, range = 24-79; mean time since stroke = 11.4 months, SD = 15.4, range = 2.0-88.0).

Methods: Spatial data were used in a 4-segment link-segment model to calculate the kinematic and kinetic variables of gait. Principal components were constructed on the averages for 40 variables.

Results: The first principal component was related to speed and accounted for 40.8% of the variance. The second principal component was related to differences between the 2 limbs (symmetry) and accounted for 12.8% of the variance. The third principal component was related to adoption of a postural flexion bias and accounted for 10.2% of the variance. The fourth principal component, which was not interpretable, accounted for 6.8% of the variance.

Conclusion and discussion: The principal-component analysis allowed clustering of related variables and simplified the complex picture presented by the large number of variables resulting from gait analysis. Examination of variables closely related to each principal component yielded insight into the nature of the strategies used in walking and their interrelationships. The method has potential for insight into similarities and differences in gait performances arising from different pathologies and for comparing the progress of individuals with similar pathologies.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Adult
  • Aged
  • Cerebrovascular Disorders / physiopathology*
  • Cerebrovascular Disorders / rehabilitation
  • Female
  • Gait*
  • Humans
  • Male
  • Middle Aged
  • Multivariate Analysis