Abstract
Factor analyses, together with the related techniques of principal components, confirmatory factor analysis, structural equation modelling, and multitrait-multimethod analysis are frequently used in quality of life (QOL) research. In particular, these methods are commonly used when exploring the ‘construct validity’ of a new QOL instrument and, to a lesser extent, to derive scoring algorithms. The current usage of factor analysis is reviewed, and problems inherent in the application of factor analysis techniques to QOL instruments are discussed. In particular,there are many examples of cases in which QOL studies
have over-used and misused factor analysis techniques.
Practical issues and problems are covered, including sample-size estimation, scaling of item-responses, and the instability of both factors and factor scores. One often overlooked issue is the impact of the definition of the population
and the selection of the samples for assessing subscale validity. The importance of this is illustrated using data from several MRC randomized clinical trials in cancer. Technical issues include the determination of the number of factors, the skewness of many QOL items, the discrete nature of many QOL items (frequently four- or five-point responses), linearity and scaling assumptions. Over recent
years there have been many advances in the methodology of factor analysis. However, most of these developments have taken place in specialised areas of psychometric test theory, and appear little known to most practitioners of
QOL studies. We review the current state of the art and discuss the controversial issues and problems that remain with factor analysis when applied to QOL data. Whilst recognizing that factor analysis can sometimes provide useful insight into the structure of a instrument, we appeal for a more cautious and restrained usage of these techniques, and for a greater recognition of the issues involved.
have over-used and misused factor analysis techniques.
Practical issues and problems are covered, including sample-size estimation, scaling of item-responses, and the instability of both factors and factor scores. One often overlooked issue is the impact of the definition of the population
and the selection of the samples for assessing subscale validity. The importance of this is illustrated using data from several MRC randomized clinical trials in cancer. Technical issues include the determination of the number of factors, the skewness of many QOL items, the discrete nature of many QOL items (frequently four- or five-point responses), linearity and scaling assumptions. Over recent
years there have been many advances in the methodology of factor analysis. However, most of these developments have taken place in specialised areas of psychometric test theory, and appear little known to most practitioners of
QOL studies. We review the current state of the art and discuss the controversial issues and problems that remain with factor analysis when applied to QOL data. Whilst recognizing that factor analysis can sometimes provide useful insight into the structure of a instrument, we appeal for a more cautious and restrained usage of these techniques, and for a greater recognition of the issues involved.
| Original language | English |
|---|---|
| Pages (from-to) | 424 |
| Number of pages | 1 |
| Journal | Quality of Life Research |
| Volume | 4 |
| Issue number | 5 |
| DOIs | |
| Publication status | Published - Oct 1995 |
