STRIEN A VAN & PANNEKOEK J (1999) Missing counts in bird monitoring programmes. LIMOSA 72 (2): 49-54.

Missing counts in monitoring schemes hamper the assessment of yearly indices and trends. Several index methods exist that cope with incomplete data. The currently most powerful method to estimate missing counts is Poisson regression (or loglinear regression). This method is available in several statistical packages and in the freeware computer program TRIM. It allows the testing and comparing of different statistical models. A simple model is the linear model by which missing counts are being estimated as if a linear trend occurs across all years. A more extensive model is the year effect model by which missing counts are being estimated from the yearly changes in other sites. The estimation of missing counts can be further improved by including environmental factors as covariates into the models. Poisson regression is suitable to deal with some other difficulties inherent in monitoring data as well, such as serial correlation, undersampling of particular strata and deviations from the Poisson distribution.

[free pdf] [dutch summary]