apple guava
New member
(but without plotting)? When I have a set of data (eg. a bunch of X values and corresponding Y values), and I plot them; it soon becomes clear to me which points are the 'weird' ones - ones that are clearly way off the line of best fit. Those dots I can ignore.
This is a decision I make visually. Firstly I identify a line or curve of best fit that passes through or closely by most points. Then I see there are a few points that are way off, and those few points I choose to ignore.
My question is: is there a way of telling which elements are the weird ones, without graphing? Assuming I don't know what the line/curve of best fit will look like, and I don't want to know what it will look like. I want some sort of algorithm (without plotting) that can tell me which elements are the least conformative to the rest. How do I do this without plotting please?
(The reason for my question is I would like a computer to be able to reject data for me, using just arithmetic, without plotting anything).
(Also the relationship between X & Y will not be the same all the time. Today in a given situation their relationship may be linear; tomorrow in another situation their relationship may be power of 3 etc.; I am after a general approach that can detect the non-conforming values and throw them out for me).
Thank you.
This is a decision I make visually. Firstly I identify a line or curve of best fit that passes through or closely by most points. Then I see there are a few points that are way off, and those few points I choose to ignore.
My question is: is there a way of telling which elements are the weird ones, without graphing? Assuming I don't know what the line/curve of best fit will look like, and I don't want to know what it will look like. I want some sort of algorithm (without plotting) that can tell me which elements are the least conformative to the rest. How do I do this without plotting please?
(The reason for my question is I would like a computer to be able to reject data for me, using just arithmetic, without plotting anything).
(Also the relationship between X & Y will not be the same all the time. Today in a given situation their relationship may be linear; tomorrow in another situation their relationship may be power of 3 etc.; I am after a general approach that can detect the non-conforming values and throw them out for me).
Thank you.