But sometimes it is true, for example in the sense that data alone does not tell much about what is being measured and its significance. For example high level

But sometimes it is true, for example in the sense that data alone does not tell much about what is being measured and its significance. For example high levels of cholesterol are negatively correlated to heart infarction, mostly because the population at highest risk is under medication and has so extremely low cholesterol. In this case the data itself is meaningless as it does not contain its own significance. This does not mean that we can freely discard data we don't like, it is important and useful, just not necessarily intrinsically meaningful. A good analysis is made so that it is obvious to justify this layer of interpretation.