作者: Nilza Nunes da Silva , Regina Bernal
DOI: 10.1590/S0034-89102009005000024
关键词: White population 、 Confidence interval 、 Landline 、 Metropolitan area 、 Statistics 、 Geography 、 Survey sampling 、 Information bias 、 Regression analysis 、 Limit (mathematics)
摘要: OBJECTIVE: To estimate landline telephone coverage effects on potential information bias in epidemiological surveys. METHODS: Databases of the Pesquisa Nacional por Amostra de Domicilios (National Household Sample Survey), from 1998 to 2003, were used rates five Brazilian geographic regions. Multiple regression analysis was identify factors associated with owning a telephone. Bias impact 95% confidence intervals assessed, according accuracy achieved each situation. RESULTS: In Southern and Central-west metropolitan regions, 70% or more, considered insignificant. remaining related above acceptable limit (0.4), indicating possible errors inferences drawn interval. Likelihood access home greater among white population higher level education. CONCLUSIONS: Findings reveal that use residential records is indicated conduct surveys states where exclusively. Specific methodologies treat estimates obtained regions lower need be analyzed published.