Assessing Psychological Need Satisfaction in Exercise Contexts: Issues of Score Invariance, Item Modification, and Context

Abstract

The researchers examined if scores from the original Psychological Need Satisfaction in Exercise Scale (Wilson, Rogers, Rodgers, & Wild, 2006) were invariant from a modified version specific to physical activity and then examined measurement invariance of scores across groups on the modified scale. Three groups were examined: (a) Students/staff from a university (N = 283), (b) a sample drawn from the general population (N = 214), and (c) individuals living with osteoporosis (N = 221). Measurement invariance was tested with four nested models using increased equality constraints per model. Results of invariance tests between two versions of the Psychological Need Satisfaction in Exercise Scale and between groups that completed the Psychological Need Satisfaction in Exercise Scale modified to physical activity supported configural and weak invariance of scores (i.e., equivalent factor structure and loadings). As such, the constructs of competence, autonomy, and relatedness were construed similarly across versions of the instrument and across two groups. Strong invariance (i.e., equivalent intercepts) was not supported, and therefore, direct score comparisons should be made with caution.

Keywords:

• self-determination theory
• psychological needs
• exercise and invariance

Acknowledgments

The first author was supported by a grant from the Social Sciences and Humanities Research Council of Canada during the preparation of this manuscript

Notes

¹The Basic Psychological Needs in Exercise Scale (BPNES: Vlachopoulos & Michailidou, 2006) was also created using the construct validation approach advocated by Messick (1995). The BPNES was developed in Greek-speaking samples of exercisers and has yet to undergo extensive validation in English speaking cultures of samples.

²A data entry error was noted for one individual's BMI value. As such, that case was deleted prior to calculating descriptive statistics for BMI (n = 252).

³These scores are based on summary composite values derived from the Godin Leisure Time Exercise Questionnaire (Godin & Shephard, 1985)

⁴Two individuals reported engaging in 100 and 240 bouts of mild, moderate and vigorous activity during a typical week. As such, these two cases were identified as extreme outliers (z > 6.00 ) and deleted prior to calculating descriptive statistics for LTPA (n = 212).

⁵Subscale scores were created using the mean of the 6 items designed to measure each psychological need. Although the data were treated as ordinal in the multi-group CFA analysis, mean subscale scores were generated for descriptive purposes only.

⁶After running the multi-group CFA analysis, LISREL applied a ridge function to the matrix being analyzed. The data were checked, and it was confirmed that the data were read into the program properly. Although this occurs frequently with polychoric or asymptotic matrices, LISREL's ridge function is an acceptable solution. One limitation is that it decreases accuracy of the robust maximum likelihood estimates.