Post
Originally shared by Christopher HanusaThe case for fitting data points visually
In my Mathematical Modeling class, we're discussing fitting data visually. One method of fitting data is by doing a linear regression, and this is a pretty automatic response to try to see what the best fit line is for some data. However, if you do this blindly, you can get into trouble because sometimes that line is not the best representation for that data. Anscombe showed this by giving four datasets that yield the same line of best fit, but as you can see in the linked blog post, it would be wrong to say that this "line of best fit" is a good fit for each data set.
My students have to learn about and write about this quartet for their next homework assignment, but I'd suggest that everyone read about it.
In my Mathematical Modeling class, we're discussing fitting data visually. One method of fitting data is by doing a linear regression, and this is a pretty automatic response to try to see what the best fit line is for some data. However, if you do this blindly, you can get into trouble because sometimes that line is not the best representation for that data. Anscombe showed this by giving four datasets that yield the same line of best fit, but as you can see in the linked blog post, it would be wrong to say that this "line of best fit" is a good fit for each data set.
My students have to learn about and write about this quartet for their next homework assignment, but I'd suggest that everyone read about it.
Anscombe's Quartet
Shared with: Public
+1'd by: stefan jeffers
This post was originally on Google+