Notes
Notes
1. The issue of recall is relevant for classification of lifetime drinking status as respondents may re-invent themselves to fit better with their current drinking identity. For example, someone who currently does not drink may reduce or even omit their past drinking when asked, particularly if this drinking occurred decades ago. Both of these issues highlight the importance of the particular question asked in determining whether, when and how much a person has drunk (Stockwell et al. Citation2004).
2. Self reported health status is particularly important to include because, of course, not all serious illnesses are immediately diagnosed and, in some populations, never diagnosed.
3. This would be critically important in general population samples in which the poor and disadvantaged are included (unlike the HPS) in that access to medical care and diagnoses are often limited.
4. Propensity score matching analyses might be utilized in this case (Rosenbauum and Rubin Citation1983). Noteworthy is that this method has recently been applied to a large sample in which it was found that moderate alcohol consumption did not “protect” against total mortality (Melberg Citation2006).
5. We add to Dr Romelsjo's observation that there is some evidence from longitudinal studies suggesting that there is a group of “virtual” abstainers (Knupfer and Room Citation1970). These are people who report abstinence but do drink on occasion (Shaper and Wannamethee Citation1998; Kerr et al. Citation2002) Of course, this group compounds the problem of potential bias in measurement particularly if the time frame is limited.
6. These include (a) consistency across several studies, (b) established experimental biological evidence of mediating processes or at least physiological plausibility (biological mechanisms), (c) strength of the association (effect size) and temporality (i.e., cause before effect) (Rothman and Greenland Citation1998).