Comparación de dos métodos en el análisis del efecto a corto plazo de la contaminación atmosférica en la salud = Comparison of two methods in the analysis of the short-term effect of air pollution on health
We compare two methods to analyse the relationship between air pollution and health. One of them (ME) is based on a generalized linear model, while the other one (MN) incorporates a generalized additive model (GAM). Besides the statistical model used, both methods schow additional discrepancies for...
| Autores: | , , , |
|---|---|
| Tipo de recurso: | artículo |
| Estado: | Versión publicada |
| Fecha de publicación: | 2003 |
| País: | España |
| Institución: | Varias* (Consorci de Biblioteques Universitáries de Catalunya, Centre de Serveis Científics i Acadèmics de Catalunya) |
| Repositorio: | Recercat. Dipósit de la Recerca de Catalunya |
| OAI Identifier: | oai:recercat.cat:10256/15748 |
| Acceso en línea: | http://hdl.handle.net/10256/15748 |
| Access Level: | acceso abierto |
| Palabra clave: | Aire -- Contaminació -- Mètodes estadístics Air -- Pollution -- Statistical methods Sèries temporals -- Anàlisi Time-series analysis Models lineals (Estadística) Linear models (Statistics) Anàlisi de regressió Regression analysis |
| Sumario: | We compare two methods to analyse the relationship between air pollution and health. One of them (ME) is based on a generalized linear model, while the other one (MN) incorporates a generalized additive model (GAM). Besides the statistical model used, both methods schow additional discrepancies for the type and number of variables used in the control of confounding. Method An analysis was carried out for each lag of black smoke (0 to 5) and several health indicators from the city of Valencia (Spain) following both methods. Results were compared examining the sequence and the lag of the higher coefficient. The ratio between confidence intervals width and the percent difference in the estimates were also obtained. Results The discrepancies in results according to the methodology were small. The sequence shaped and the lag of higher coefficient generally remained invariant. Coefficients were similar (percent difference of MN over ME was not greater than 10%). Confidence intervals were more precise for MN, although the reduction was not very strong (the width ratio was around 90%). Conclusions The short variation of the results guarantees the applicability of both MN and ME methods, although the systematic reduction of the standard error of MN supports its use as a more efficient procedure to control for confounding |
|---|