 Good morning, everyone. Thanks for being here. We present the party package that permits to perform a valid double dipping. And I will use, as example, deprimary data that are very important in order to understand the neural activation in the human brain. One of the most famous methods used in deprimary data analysis is the cluster-wise method. In this case, we analyze a set of continuous voxel, S. And for each of these set, called cluster, we have analysis in order to understand if the brain activation inside that cluster is significant. However, if we reject this analysis, we only know that at least one voxel is active. But we don't know how many and which ones. This is called a specificity paradox because larger the cluster, weaker the findings. And we cannot infer inside that cluster without falling into the double-dipping problem. So in this case, we can use the ARRI method that stands for a resolution inference or the party method that is the permutation version of ARRI. That these two methods permits to make inference on the number of truly active voxels. So we can associate it to each cluster the lower bound of the number of truly active voxel. And the permutation structure of ARRI permits to account for the correlation structure between tests. And both methods are based on the closing testing procedure for controlling the family-wise error rate. So every time that you want to infer inside a data-driving cluster and not, you can use ARRI, for example, in gene expression cluster analysis, in cluster AG data analysis, and so on, as many times as you want because it's a simultaneous inference. And so I will present the main function of the package, the party bring function and party function. The first one is for the fMRI framework and the second one is for the general framework. In the first case, we need the object cops that is a list of contrast parameter estimate involving bring activation difference for each subject in if the format. We need the detraction that constructs the clusters or the cluster map, the brain mask, and the alpha level. And for the general framework, we need the data matrix with dimension observation times variable, the feature set of interest, so the set of hypothesis, the alpha level, and the type of test that we want to perform. So for example, using the auditory data that you can find in the fMRI data package, in this case, we want to analyze the neural activation between vocal and vocal stimuli. And so using these three objects and using the party bring function, we can have an output like that. So the first line is for the first cluster that is the right superior temporal gyrus, plant temporale, and so on. This cluster comes from the cluster-wise method using that ratio equal to 3.2 for data statistics that form the clusters. And we can say that in this first cluster, we have at least 92% of truly active voxel. But we can also drill down. And so we can use a higher duration, for example, equal 4. And in this case, we have four subclusters inside that cluster. And we can compute the lower bound of the number of truly active voxel for each of these subclusters. We can also visualize the results in a brain map using the mapTDP function. So we can see, for example, in this case, we have a higher true discovery proportion respect to this cluster. And that's it. So I hear first question. And I want to thank the group that work on this project, the Professor Finos, Goehmann, Hamerick, and Bida. And thanks for your attention.