Because my job is very much science and research oriented, a huge part of what I do on a daily basis, aside from simply trying to avoid getting fired, is to evaluate the quality of research. Simply because something is published, doesn't tell you anything about the level of quality. One example of such a concept was brought to my attention as I browsed Fark.com this afternoon. See here.
Essentially, the research, which appears in the Annals of Behavioral Medicine, says that people who attend church have better pulmonary flow rates than those who do not. Pulmonary flow rate is one measurement that is thought to correlate with better health in old people. Wow. Guess we all better start going to church right?
While on the surface this seems encouraging to those who attend church services, this is a great example of terrible research, simply because the researchers are over-interpreting their findings. Correlation, my friends, is a very very different thing than causation. While the study does show that the church attendance in people 70-79 years old is linked to better pulmonary health, that does not mean one can infer that church is what caused those people to have better health.
The fact that a person is in there 70s and is ABLE to get up and go anywhere is probably a greater predictor of pulmonary function than where they end up going. A better way to look at this question would be to include individuals in the same age group that are matched for age and activity-level who are athiests to those church-going grannies. Something tells me the two groups would be very similar.
Looking into this further, I discovered that another group of researchers published a similar study earlier this year, that tried to conclude that church attendance would add years on to your life. Again, the data don't support their conclusions. While it may be true that people who go to church do live longer on average, the causal relationship between church attendance and longevity is suspect.
This is why good researchers use control groups that are matched for age and other significant factors, because otherwise, you can't and shouldn't make causal conclusions based on correlational research.