ATAR Notes: Forum

VCE Stuff => VCE Mathematics => VCE Mathematics/Science/Technology => VCE Subjects + Help => VCE General & Further Mathematics => Topic started by: geoff_821 on November 06, 2011, 09:58:22 pm

Title: Comparing LSRL with 3-median/mean line
Post by: geoff_821 on November 06, 2011, 09:58:22 pm

When would the Least Squares Regression Line, Three Median Line and Three Mean Line be appropirate to use, and why is one better than the other.

I know we need to use a three median line when there are "extreme values" present because it won't get influenced.
What about the others?

Title: Re: Comparing LSRL with 3-median/mean line
Post by: thatisanote on November 06, 2011, 10:35:38 pm

don't think we need to compare the others (other than 3-med and least squares) but I'm interested to hear the thoughts of others.

3 median line is *less* influenced by outliers, though it can be influenced. (if the outlier is fixed up as it was an error in the data, it is possible that the median of a section could change)

Title: Re: Comparing LSRL with 3-median/mean line
Post by: smithy16 on November 06, 2011, 10:40:05 pm

In what type of question would you say that one of the values found is an error? Least Squares or??

Title: Re: Comparing LSRL with 3-median/mean line
Post by: thatisanote on November 06, 2011, 10:54:14 pm

Quote from: smithy16 on November 06, 2011, 10:40:05 pm

In what type of question would you say that one of the values found is an error? Least Squares or??

nah you wouldn't say it but vcaa might.
e.g. it has the plot.
then it says 'it turns out that this value was an error and was actually meant to be xyz'

is it still an outlier?
or something like that