A majority of quasar surveys have been based on criteria which assume strong blue continua or a UV-excess. Any amount of dust along the line-of-sight is expected to drastically extinguish the optical/UV flux leading to a selection bias. Radio surveys however should suffer no bias against extinction by dust. Recently, a large complete sample of radio-selected quasars has become available (the `Parkes sample'). A majority of these sources exhibit optical-to-near-infrared continua that are exceedingly `red', very unlike those of quasars selected optically. The purpose of this thesis, broadly speaking, is to explore the problem of incompleteness in optical quasar surveys due to obscuration by dust, and to interpret the relatively `red' continua observed in the Parkes quasar sample.
The first part of this thesis explores the observational consequences of an intervening cosmological dust component. A preliminary study explores the effects of different foreground dust distributions (on galaxy-cluster scales to the visible extent of normal galaxies) on obscuration of background sources. Numerical simulations of dusty-galaxies randomly distributed along the line-of-sight with simplified assumptions are then performed, and implications on optical counts of quasars and absorption-line statistics are explored. This foundation is extended by considering the effects of more complicated models of foreground obscuration where the dust content evolves with redshift. The Parkes sample is used to constrain evolutionary and physical properties of dust in intervening systems. The contribution of line-of-sight galactic dust to the reddening observed in this sample is also constrained.
The second part examines the continuum properties of Parkes quasars in the framework of a number of absorption and emission mechanisms to assess the importance of extinction by dust. Three classes of theories are explored: `intrinsically red' AGN emission models, dust extinction models, and host-galaxy light models. Simple models are developed and tested against the available data. Several new correlations between spectral properties are predicted and identified observationally. For the dust model, we explore the effects of dust on soft X-rays and compare our predictions with ROSAT data. Possible physical dust properties are constrained. I then consider the possibility that a `red' stellar component from the host galaxies contributes to the observed reddening. This contribution is quantified using a novel spectral fitting technique. Finally, an observational study of near-infrared polarisation is presented to distinguish between two models for the reddening: the intrinsic `synchrotron emission model', and the dust model. Combined with spectral and photometric data, these observations are used to constrain various emission and dust absorption models.