School of Engineering and Computer Science Te Kura Mātai Pūkaha, Pūrorohiko

Here's a rough guide to Statistics, which we'll expand over time:

• Significance Testing for Classification.pdf: Signifcance slides from Ben Evans

In terms of understanding the statistics, please have a look at this comic:

https://xkcd.com/882/

it's the best illustration of familywise errors that we have come across! Please make sure you can understand and explain the difference between type I and type II errors. Then understand what happens when you apply multiple tests on the same sample e.g. compare your method against one of a number of different methods on the same sample (dataset). Then consider whether you need to make a correction, e.g. a Bonferroni correction [you may not need to make a correction, but you need to understand that you are in danger of accidentally making a type I error].

This is a really good resource:

http://www.how2statsbook.com/

where this is a really practical video:

https://www.youtube.com/watch?v=kpTvTyMcDqY

This looks a useful grid for choosing a statistical test

If you wish to compare a single value to a distribution then worth considering a one sample t-test

btw this explains the survival analysis column in the grid above

Comments more than welcome