6.1.2 Selection Effects at Work

Selection effects have to be considered in comparing objects drawn from different surveys and for the derivation of a luminosity function, as first pointed out by Wasilewski (1983) and summarized by Comte (1998), who noted that selected populations have different statistical properties that depend of their selection modes. Emission line selected samples span a broader range of colours than purely UV-excess objects. This is not a surprise. Salzer et al. (1997a) pointed out that very blue objects selected from their emission lines were also missed by the Markarian surveys. Similarly the distribution of emission line equivalent widths strongly depends on the choice of the dispersion. Very young starbursts are favoured in emission line surveys while ageing bursts are more easily picked out in ``continuum'' surveys. Finally H surveys present no prejudice in favour of a given metallicity.