Return to Video

vimeo.com/.../722088590

  • 0:06 - 0:09
    Confounding
    variables are a type of variable
  • 0:09 - 0:13
    related to a model's independent
    and dependent variables.
  • 0:13 - 0:17
    A variable must meet two conditions
    to be a confounder.
  • 0:17 - 0:21
    One, it must be correlated
    with the independent variable.
  • 0:22 - 0:27
    And two, it must be causally related
    to the dependent variable.
  • 0:27 - 0:29
    A pretty classic example
  • 0:29 - 0:35
    tasks you with collecting data on sunburns
    and ice cream consumptions.
  • 0:35 - 0:37
    If we find that ice cream consumption
  • 0:37 - 0:41
    goes up at the same time
    as instances of sunburn,
  • 0:41 - 0:45
    are we meant to conclude
    that ice cream causes sunburn?
  • 0:46 - 0:51
    One missing variable here is temperature
    as a proxy for the amount of sun
  • 0:51 - 0:53
    as a confounding variable.
  • 0:53 - 0:56
    The hotter
    it is, the more likely people are to eat
  • 0:56 - 1:01
    ice cream, as well as get a sunburn
    as they spend more time outdoors.
  • 1:01 - 1:06
    One of the challenges of modeling is
    that it is necessary to do an exhaustive
  • 1:06 - 1:11
    search for possible confounding factors,
    as their absence could lead to
  • 1:11 - 1:16
    certain algorithms to detect relationships
    that don't actually exist.
  • 1:16 - 1:20
    By the same token,
    if the confounding factor is identified,
  • 1:20 - 1:23
    but not properly handled
    by the algorithm in question,
  • 1:24 - 1:28
    they could end up
    exaggerating relationships that do exist.
Title:
vimeo.com/.../722088590
Video Language:
English
Duration:
01:30

English subtitles

Revisions Compare revisions