Correspondence analysis is a statistical method developed for descriptive exploratory analysis of large matrices. Through this technique, the most important relations between a large pool of categorical and continuous variables may be plotted graphically. In this paper the described technique was applied to the data of an evaluation process of the National Program of Immunizations (PNI). This evaluation was carried out in 660 health clinics, distributed over 98 Brazilian cities. The variables analysed were related to the performance in PNI activities or potential determinants of this performance. The correspondence analysis was followed by an ascending hierarchical classification process of the Health Clinics, using the computer program "Systeme Portable Pour L'Analise de Données - SPA". The performance pattern was not homogeneous. The clinics were classified in five groups according to the performance in the PNI activities. In 58% of the Health Clinics, even the most elementary norms for the PNI were not followed. The correpondence analysis is a powerful tool for inspecting large data matrices and selecting the most important variables for a more detailed subsequent analysis.
Correspondence Analysis; Assessment; Vaccination
