I have calculated mean, std. deviation, IQR, trim mean, max/min eigen values, etc for each feature of the dataset and took an average of each of the measures to get a single value for each of the measures for one dataset. But these are usually used for univariate data. I was wondering what would be a better measure for multivariate data which will give me a single value to represent a whole dataset.

I will later use these measures to compare multiple datasets.