prev next front |1 |2 |3 |4 |5 |6 |7 |8 |9 |10 |11 |12 |13 |14 |15 |16 |17 |18 |19 |20 |21 |22 |23 |24 |25 |26 |27 |28 |29 |30 |31 |32 |33 |34 |35 |36 |37 |38 |review
These probabilistic models are very useful. They help us to make a lot of decisions we could not easily make otherwise. But it is important to remember that these are just probabilities. There will always be people who do not have the problem, even with extreme values. This is just like the probability of winning the Mark Six. The chances are extremely low, but someone wins more or less every week. The problem isn't knowing whether someone will or will not win, but exactly who and when. Probabilistic models cannot answer this latter point very well.

Probabilistic models do not easily handle extreme chance scores, nor do they transfer well to other systems. Some systems are inherently more variable than others, and some tests are more accurate than others. This view also doesn't comfortably allow estimates of such important data as disease attack rates or the extent of treatment benefits.

To help overcome some of these shortcomings, epidemiologists use two measures of tests: sensitivity and specificity, to help determine how valuable a test will be in predicting who will develop the disease (Positive Predictive Value - PPV) and who will not.

Don’t worry about these now. You will not be asked to calculate PPVs or sensitivity and specificity of tests, but you need to be aware that these concepts are important in deciding how useful a test is.