Python Data Mining Quick Start Guide
上QQ阅读APP看书,第一时间看更新

Basic summary statistics

Practitioners in the field of descriptive analytics use a set of four summary statistics to quickly understand a dataset. With practice, you should be able to strengthen your intuition about each one of these statistical measurements. In fact, it's a great place to start with most problem statements that you will face. The four summary statistics are described as follows: 

  • Locations: The location or center of the data; this can be measured by the mean (average), median, or mode. The median is the point of delineation in 50% of the data, and the mode is the most occurring points, or largest part of the distribution. 
  • Spread: How the data is spread around the center; this can be measured with standard deviation, which sums the average distance from the mean of each data point, or variance, which is the square of the deviation. 
  • Shape: A description of where the center of distribution sits in relation to the mean. This is usually expressed as the skew direction. You can refer to the following diagram for a negative skew example. In the case of positive skew, the tail is simply pointed in the opposite direction. 
  • Correlation: The measurement of dependency of one variable against another. The most common measure is the Pearson correlation coefficient, which is between -1 (a full negative correlation) and +1 (a full positive correlation). A value of 0 signifies no correlation; this is usually denoted with "r". 

Take a look at the following diagram for a visualization of the points described in this section: