Quantiles

quantile(matrix: any[] | Mat | Tensor, q: number) : any[]

param matrix - an Array, Mat or Tensor - where the data is stored to find the quantiles of

param q a number in the interval [0, 1]

returns - any[] - an Array representing the quantile (all the numbers that fall below the quantile limit)

The quantile is a descriptive statistic. Note that it is similar to the percentile, or percentile(), another function in this book. The quantile is a number between 0 and 1 that cuts the distribution of points or values in an Array evenly (by numerical order) and labels them with their appropriate quantile. For example, taking all the Array's data in order, splitting it in half evenly, and taking the middle point is quantile 0.5. Hence, iqr(interquartile range) is also related to quantiles.

Quartiles are quantiles cut at 1/4 parts - 0.25, 0.50, 0.75. The other quantiles are arbitrary. Overall, if you want to find the bottom y% of data, you can simply call quantile(matrix, y/100). That is because y/100 converts a percentage value into a quantile value. For example, if I want the bottom 23% of values in my 2-D Array called matA, I would simply call quantile(matA, .23). This gives me the .23 quantile of the input, an Array with all values below 0.23.