Validity & Reliability: Introduction

00:01 Hello and welcome to epidemiology. Let me ask you question, how often do you use the word 'cause' in a clinical context, like smoking causes cancer, or eating fatty food causes heart disease? You probably do it a lot. Did you ever stop to think about what cause really means. Epidemiologists are quite particular about the nature of causality or causation, certain criteria, certain conditions have to be satisfied before we're confident using the word 'cause', as in this risk factor causes a certain outcome. So in today's lecture you're going to learn about a set of criteria we' like to apply to causation, we call them Bradford-Hills criteria. You will also learn about the difference between necessary and sufficient causal factors, that's an exercise in logic that applies to science. And lastly you're going to learn the difference between validity and reliability, classical constructs in scientific proof.

00:52 Now let's look at this graph. This shows the relationship between the consumption of ice cream and the number of drowning deaths in a New England community. Now when we look at it, it shows that when on days in which consumption of ice cream goes up, the number of drowning deaths also go up. So what do you think? Is consumption of ice cream a causal factor in drowning deaths, in other words, if you eat ice cream, you're going to drown.

01:17 Is that what this is saying? No, because correlation is not causation, I think you know that. Oftentimes we can measure associations and relationships between variables; it doesn't mean that one of them is causing the other one. What's actually happening in that example is that on warm days, people are more likely to eat ice cream and they're more likely to go to the beach and probably drown. So in that case, if you remember the lecture on biases, temperature is a confounder. Temperature is creating an artificial relationship between two variables that actually don't have a relationship, in this case ice cream consumption and drowning.

01:55 So in a previous lecture we've also learned about measurements of association, in particular relative risks, odds ratios and attributable risk. We never used the word causal though, we're very careful to say these are measurements of association or relationships, because to say that one thing causes something else is another story entirely and that's what we're going to talk about now.