How to get average, median, mode value in BigQuery

Share this for your friends.

How can we get average, median, mode value in BigQuery ?

BigQuery(BQ) is very useful for data analysis or processing.

It is good at handling huge data. It returns summary result in short time.

And we want average, median, mode value when we analyse data.

How can we get these values in BigQuery ?

So today I introduce about "How to get average, median, mode value in BigQuery".

Author

Mid-carieer engineer (AI, system). Good at Python and SQL.

Advantage to read

You can understand "How to get average, median, mode value in BigQuery".

Data

First, we have to prepare data.

Load this CSV file into table avg_median_sample.

text_col,num_col1,num_col2,num_col3
a,10,1,100
a,10,2,100
a,10,3,100
a,10,4,100
a,10,5,100
a,10,6,100
a,10,7,100
a,10,8,100
a,10,9,100
a,10,10,100
b,12,11,200
b,12,12,200
b,12,13,200
b,12,14,200
c,15,15,200
c,15,16,200
c,15,17,300
c,15,18,300
c,15,19,10000

Then we can see table like below.

Row	text_col	num_col1	num_col2	num_col3
1	a	10	1	100
2	a	10	2	100
3	a	10	3	100
4	a	10	4	100
5	a	10	5	100
6	a	10	6	100
7	a	10	7	100
8	a	10	8	100
9	a	10	9	100
10	a	10	10	100
11	b	12	11	200
12	b	12	12	200
13	b	12	13	200
14	b	12	14	200
15	c	15	15	200
16	c	15	16	200
17	c	15	17	300
18	c	15	18	300
19	c	15	19	10000

How to get average value

Once data is prepared, get average value.

In order to get average value, you should use AVG() function.

SELECT
text_col,
avg(num_col1) as avg_1,
avg(num_col2) as avg_2,
avg(num_col3) as avg_3
FROM test.avg_median_sample
GROUP BY text_col

Result is like below.

Row	text_col	avg_1	avg_2	avg_3
1	a	10	5.5	100
2	b	12	12.5	200
3	c	15	17	2200

Reference

Expressions, functions, and operators in Standard SQL | BigQuery

How to get median value

Next, get median value.

Median value is not a calculated value.

It is middle value in ordered list.

What is difference between average and madian ?

Average value is the center of gravity. If data scale is same in data list, it is good feature.

But if there is big scale value in data list, average value also become large. In that case, it is not a good feature.

Umm, any example ？

For example, imagin that there are 100 people and Optimus Prime. What does average value mean ?

Oh, Optimus Prime is too heavy. So average value is useless.

We can get median value with PERCENTILE_CONT() function.

We use it as below.

SELECT distinct
text_col,
PERCENTILE_CONT(num_col1, 0.5) OVER(PARTITION BY text_col) AS median_1,
PERCENTILE_CONT(num_col2, 0.5) OVER(PARTITION BY text_col) AS median_2,
PERCENTILE_CONT(num_col3, 0.5) OVER(PARTITION BY text_col) AS median_3
FROM test.avg_median_sample
ORDER BY text_col

Then we get this result.

Row	text_col	median_1	median_2	median_3
1	a	10	5.5	100
2	b	12	12.5	200
3	c	15	17	300

Average value of text_col="c" and num_col3 was 2200.

Median value is 300. So we can understand this difference.

And if you change second parameter of PERCENTILE_CONT() from 0.5 to 0.1 , you can get 10% position value. If 0.75, you can get 75% position value.

Reference

Expressions, functions, and operators in Standard SQL | BigQuery

How to get mode value

Then try to get mode value.

The mode is the value that appears most frequently in a data set.

Frequency is important. 100 and 100.1 are different values.

So mode doesn't match for real numbers.

It is good to use for checking group frequency.

In order to get mode value, you vcan use APPROX_TOP_COUNT().

SELECT
text_col,
APPROX_TOP_COUNT(num_col1, 1) top1,
APPROX_TOP_COUNT(num_col2, 1) top2,
APPROX_TOP_COUNT(num_col3, 1) top3
FROM test.avg_median_sample
GROUP BY text_col
ORDER BY text_col

Result is below.