16.20 - Option 3: Replace Missing Values with an Estimated Value - Teradata Database - Teradata Vantage NewSQL Engine

Teradata Vantageā„¢ Time Series Tables and Operations

Product
Teradata Database
Teradata Vantage NewSQL Engine
Release Number
16.20
Published
March 2019
Content Type
Programming Reference
Publication ID
B035-1208-162K
Language
English (United States)
Last Update
2019-05-02

You can estimate a value to update the rows with a missing value.

The value loaded can be the result of any function, including other aggregates.

The example shows this for the TEMPERATURE column in Table and Data Definition for Time Series Aggregates Examples. The example updates all missing values in each timebucket with the average of the present values:

Create table ocean_buoys2 as ocean_buoys with no data;
INSERT INTO OCEAN_BUOYS2 VALUES(TIMESTAMP '2014-01-06 08:00:00.000000', 0, 55, );
INSERT INTO OCEAN_BUOYS2 VALUES(TIMESTAMP '2014-01-06 08:09:59.999999', 0, 55, );
INSERT INTO OCEAN_BUOYS2 VALUES(TIMESTAMP '2014-01-06 09:01:25.122200', 1, 55, );
INSERT INTO OCEAN_BUOYS2 VALUES(TIMESTAMP '2014-01-06 09:02:25.122200', 1, 55, );
 
MERGE INTO OCEAN_BUOYS2
USING (SELECT TD_TIMECODE, BUOYID, Avg(TEMPERATURE) FROM OCEAN_BUOYS GROUP BY (TD_TIMECODE, BUOYID) WHERE TEMPERATURE IS NOT NULL) AS S(a,b,c)
ON TD_TIMECODE=S.a AND BUOYID=S.b AND TEMPERATURE IS NULL
WHEN MATCHED THEN UPDATE SET TEMPERATURE=S.c;
select * from ocean_buoys2 order by 1;

Result:

TIMECODE BUOYID SALINITY TEMPERATURE
2014-01-06 08:00:00.000000 0 55 10
2014-01-06 08:09:59.999999 0 55 99
2014-01-06 09:01:25.122200 1 55 73
2014-01-06 09:02:25.122200 1 55 74
SELECT $TD_TIMECODE_RANGE, $TD_GROUP_BY_TIME, BUOYID, AVG(TEMPERATURE), COUNT(*)
FROM OCEAN_BUOYS
WHERE TD_TIMECODE BETWEEN TIMESTAMP '2014-01-06 08:00:00' AND TIMESTAMP '2014-01-06 10:30:00'
AND BUOYID=0
GROUP BY TIME (MINUTES(10) AND BUOYID)
ORDER BY 2,3;

Result: Note the effect on the Average column.

TIMECODE_RANGE GROUP BY TIME(MINUTES(10)) BUOYID Average(TEMPERATURE) COUNT(*)
('2014-01-06 08:00:00.000000+00:00', '2014-01-06 08:10:00.000000+00:00') 1 0 54 3
('2014-01-06 08:10:00.000000+00:00', '2014-01-06 08:20:00.000000+00:00') 2 0 55 2