A.C. Pronk You can assess the outliers through Z-Score and verify, if the outliers are due to data-entry errors and remove them if they are erroneous or not relevant. If they are relevant, you can transform them into logs or square roots. After the move, you can conduct an analysis of the effects on your results.

A.C. Pronk -

I see that you used graphs to explore your data. I agree with that approach. Often graphics can help analyze the overall context as well as see what does not fit. There may be a pattern that helps you analyze the problem, or perhaps you could have one point where it is obvious that a number was reported in the wrong units. Regardless, you do not simply throw out a point because it is inconvenient. It may be relevant. Further, points that don't belong may or may not be extreme cases. Nothing should be thrown out without an analysis as to the possible causes for a point to appear to be an outlier.

Simple graphics are often useful, such as comparing current to past values. At any rate, graphics, particularly scatterplots, are often a valuable tool in various applications.

Cheers - Jim

If you have assumed outliers in your data: