A family of Microsoft spreadsheet software with tools for analyzing, charting, and communicating data.
Everybody from the respondents here could have proposed the same method as your brother did. There is only the question, whether this would be able to give the public true image of reality, as you had certainly intended. A regression line would be able to represent correct statistical treatment in average; it can be completed with standard error lines. Then the swarm of original points could optically reflect the character of scatter. A danger can occur that just the eliminated points could be those that draw the statistical parameters in an unexpected direction; this is a nuisance of solitary distant points. And this is the reason, why you were advised to use different size markers (I wonder why there is a simple macro on the net). Of course, the truth is that this danger is the lower the higher is the count of points. Your proposal to use only 1 out of 250 points suggests that you have to do with ten thousands; perhaps this would be OK. For a reasonable number it is sufficient to depress the marker + size into minimum 2 to 3 points.
Petr Bezucha