More Summary Functions

Exploratory Data Analysis in SQL

Christina Maimone

Data Scientist

Correlation

Scatterplots of data points with positive and negative correlations

Correlation function

SELECT corr(assets, equity) 
  FROM fortune500;

       corr        
-------------------
 0.637710143588615
(1 row)

Median

1 1 4 4 4 5 6 7 13 19 20 20 21 21 22
              ^
            median
        50th percentile

^                                  ^
0th percentile      100th percentile

Percentile functions

SELECT percentile_disc(percentile) WITHIN GROUP (ORDER BY column_name)
  FROM table;

-- percentile between 0 and 1

Returns a value from column

SELECT percentile_cont(percentile) WITHIN GROUP (ORDER BY column_name)
  FROM table;

Interpolates between values

Percentile examples

SELECT val
  FROM nums;

 val 
-----
   1
   3
   4
   5
(4 rows)

SELECT percentile_disc(.5) WITHIN GROUP (ORDER BY val),
       percentile_cont(.5) WITHIN GROUP (ORDER BY val)
  FROM nums;

 percentile_disc | percentile_cont 
-----------------+-----------------
               3 |             3.5

Common issues

Error codes
- Examples: 9, 99, -99
Missing value codes
- NA, NaN, N/A, #N/A
- 0 = missing or 0?
Outlier (extreme) values
- Really high or low?
- Negative values?
Not really a number
- Examples: zip codes, survey response categories

Let's practice!

Exploratory Data Analysis in SQL