Hi Professor, Let's take Afghanistan as an example. In the sample2.tsv file, there are 9 records, ignoring 2013 and 2014. I assume this is because of the empty values for the homicide rate in those two years. Then we use these 9 records to calculate correlations. As you mentioned in the clarification emails, "this may correspond to empty cells in one of the predictor columns, and the number of correlations may be slightly different for each predictor." Based on this condition, I assume the two years should be considered valid. I am unsure about whether all three predictors need to have values (inner join) when we join the three files, or if it is allowed to have empty values (outer join). Thanks for your clarification!