In statisticiansix/demonstrandum:

Association introduction

Measures of association look to quantify the strength of a relationship between two variables within a data set, the most well known of these is correlation and is often used as a catch-all term. The way in which we calculate the measure of association is dependent upon the type of data that we wish to analyse (continuous, integer, nominal, ordinal, or categorical), and for each of these should the value be 0 there is no association and should the value be 1 this indicates perfect associatio, values between 0 and 1 indicate the strength of the association - with some tests the value can be negative and this indicates the how the variables respond to changes in each other.\n There is a saying when it comes to measures of association which is \"correlation does not imply causation\" and this is true, just because two variables are associated does not mean that the changes in one are caused by another. An example of this would be INSERT EXAMPLE HERE. Whilst the two variables are associated there is a third, potentially unknown, variable that is causing the change in the variables we have analysed. If one variable causes a change in value in the other this is called a \"causal\" relationship and cannot be identified solely using data analysis, it requires a wider understanding of where this data has come from. Another example is where two variables are associated but there is no hidden variable such as the length of the winning word in the Spelling Bee is associated with the number of deaths from venomous spiders. This has no possible causal relationship but is associated and in such cases these relationships are called \"spurious associations\", this example came from Spurious Correlations (http://tylervigen.com/spurious-correlations) which has plenty of examples of spurious associations. We have been referring to association to describe relationships rather than correlation as correlation is just one type of association that we are going to use.

Descriptions

Detailed Descriptions

As with most statistical methodologies traditional correlation has a set of underlying assumptions that the data being analysed must meet, and if these are not true then we have alternatives that we can use instead.

statisticiansix/demonstrandum documentation built on Dec. 2, 2019, 1:29 a.m.