ukb-extract

Usage

    1. Open extract.py and set paths and user-options
    2. python extract.py

Data-field names in UKB file are coded as

f . field-id . visit number . instance number

ex) f.34.0.0

They can be one of the following types:

• Integer
• Categorical (single)
• Categorical (multiple)
• Continuous
• Text
• Date
• Time
• Compound

extract.py

1. Webscrape descriptions, categories and variable-types for each field-id from: http://biobank.ctsu.ox.ac.uk/crystal/list.cgi.
2. Enumerate field-id’s for each variable type.

    * Output/Integer.datafields
    * Output/Categorical_single.datafields
    * Output/Categorical_multiple.datafields
    * Output/Continuous.datafields
    

3. Split UK Biobank by variable types: Integer, Categorical and Continuous.
   • For continuous and integer features, encode missing values as ‘NA’.

       * Output/Integer.phenoslice.csv
       * Output/Continuous.phenoslice.csv
       

   • For categorical features,
     • Enumerate all possible categories.
     • One-hot encode all field-id/category combinations.

           * Output/Categorical_single.phenoslice.csv
           * Output/Categorical_multiple.phenoslice.csv
       

4. Merge dataset into a single file.
   • Output/features_merged.csv

Note:

• ‘f.xxxxx.x.x’ represents integer or continuous columns
  • ex) f.3064.0.1 would represent the measurement #1 of data-field 3064 on assessment day 0.
• ‘c.xxxx.x’ represents categorical columns.
  • ex) c.3606.1 would be the indicator that an individual is in category 1 (‘Yes’) for data-field 3606.

To look up the significance of a field, use the UK Biobank data showcase: http://biobank.ndph.ox.ac.uk/showcase/