[MUSIC]. Robustness is about finding the same results, whether correct or not correct. Now we want to be sure that we classified individuals in the correct box. True positive, false positive, negative, and so on. This is about validity. We want to be correct, which is not the priority when you use robustness. Validity in general is an estimate that is equal to the parameter you are trying to find. Usually, in validity, we will talk about sensitivity and specificity. To assess the sensitivity and specificity of a test, you will have to compare two tests, or at least two tests. The one for which you want to assess is performance, its performance would be sensitivity, specificity, compared to a test what we call the gold standard or, more generally, the usual care. What you use in general in the clinical practice. A new test compared to another test. If you have discrete measures, high blood pressure, low blood pressure, if you have high risk of disease, low risk of disease, you will, in general, use sensitivity and specificity. You have almost dichotomous measures. You could put someone into a two-by-two table we see in epidemiology. If you use continuous measures, then you cannot put people on the cells, but rather, you would use correlation as a marker or a measure of validity. So let's discuss sensitivity and specificity. Usually, as you can see on this slide, sensitivity and specificity are discussed in a two by two table. The test, which is positive or negative, based on the cutoff you choose, and the disease, which is the presence of the disease or the absence of the disease. In general, people are located in one of the four cells. They might be A, B, C, or D. Sensitivity specificity will help you describe or assess the number of people, what we call true positive, you know he has the disease and he tests positive, this is the TP cell. Other participant are known not to have the disease, and you're happy because your test is negative. This will be the true negative. When there's a mismatch between the disease and the results of the test, you will end up having false negatives, people that you know have the disease but test negative. And the false positive, people who you know do not have the disease, but yet test positive using your test. Most of the time, sensitivity specificity is a ratio of the « a » cell on the « c » plus « a » cell. Yet I would rather suggest you to remember this formula which is saying that the sensitivity is equal to the, I would say, Th proportion of probability of having a positive test given that you have the disease. This is the mathematic expression of this definition. The probability or proportionality of having a positive test, T +, given that you have the disease. The specificity would be the probability of having a negative test given that you don't have the disease. The given is really important, because as you can see on this slide, it really helps you understand what is, by definition, given. In sensitivity and specificity you know who is or who has the disease. You kind of master that usually taking rats, animal models, or using with the disease and without the disease, and you try to determine how good your test is to differentiate people with and without the disease based on what is given, the disease. We'll see, using preventive value, that in clinical practice, we don't use test, after knowing that you have the disease, there's no way or no reason, it doesn't make sense to assess the performance of a test for a patient saying I know that you are sick, let's try how good the test is. We want the test to determine the probability of having the disease. This is what we call the preventive, or the predictive, value. [MUSIC]