Detecting inconsistent data

Cleaning Data in PostgreSQL Databases

Darryl Reeves, Ph.D.

Industry Assistant Professor, New York University

Inconsistent data

Certain restaurant inspection rules

score corresponds to number of violations 2_4_hazard

camis    |              name               |  score | ... 
---------+---------------------------------+--------+-----
...      | ...                             | ...    | ...                                   
41659848 | LA BRISA DEL CIBAO              | 20     | ...
40961447 | MESON SEVILLA RESTAURANT        | 50     | ...
50063071 | WA BAR                          | 15     | ...
...      | ...                             | ...    | ...

A (0 to 13), B (14 to 27), C (28+)
Scenarios for grades:
- A on initial inspection
- Re-inspection with A, B, or C

¹ https://www1.nyc.gov/assets/doh/downloads/pdf/rii/restaurant-grading-faq.pdf

Checking rules with SQL

Interdependent can introduce inconsistency
Rules can be encoded in SQL
A given for score from 0 to 13

SELECT 
    camis, 
    grade, 
    grade_date, 
    score, 
    inspection_type 
FROM 
    restaurant_inspection 
WHERE 
    grade = 'A' AND 
    score NOT BETWEEN 0 AND 13;

Checking rules with SQL

B given for score from 14 to 27

SELECT 
    camis, 
    grade, 
    grade_date, 
    score, 
    inspection_type 
FROM 
    restaurant_inspection 
WHERE 
    grade = 'B' AND 
    score NOT BETWEEN 14 AND 27;

 camis    | grade | grade_date | score |         inspection_type          
 ---------+-------+------------+-------+----------------------------------
 50034653 | B     | 12/06/2019 | -1    | Cycle Inspection / Re-inspection

Checking rules with SQL

SELECT 
    camis, grade, grade_date, score, inspection_type FROM 
    restaurant_inspection 
WHERE 
    (grade = 'A' OR grade = 'B' OR grade = 'C') AND
    inspection_type LIKE '%Reopening%';

 camis    | grade | grade_date | score |                 inspection_type                 | ... 
 ---------+-------+------------+-------+-------------------------------------------------+-----
 ...      | ...   | ...        | ...   | ...                                             | ...
 50005784 | C     | 05/29/2019 | 14    | Cycle Inspection / Reopening Inspection         | ...
 50091190 | C     | 07/12/2019 | 7     | Pre-permit (Operational) / Reopening Inspection | ...
 40395023 | C     | 09/13/2019 | 8     | Cycle Inspection / Reopening Inspection         | ...
 50037770 | C     | 10/26/2018 | 11    | Cycle Inspection / Reopening Inspection         | ...
 50036406 | C     | 07/10/2018 | 20    | Cycle Inspection / Reopening Inspection         | ...
 ...      | ...   | ...        | ...   | ...                                             | ...

¹ https://www1.nyc.gov/assets/doh/downloads/pdf/rii/restaurant-grading-faq.pdf

Data cleaning insights

Diversity of approaches
Careful thought required
Domain knowledge is key
- Which values are valid
- Reasons for duplication
- Appropriate fill-in values

Let's practice!

Cleaning Data in PostgreSQL Databases