Association and causality

how do we analyze whether one variable is causally related to another

• a relationship must satisfy 3 criteria to be considered causal

  • association

  • time order

  • elimination of alternative explanations

• Association:

  • shows that x and y are associated as x changes the distributions of y should change in some way

  • some ways to test this is the:

    • chi squared test

    • t-test or f test for regression slope or for comparissions of means

• time order

  • sometimes it’s just a matter of logic

  • In other cases, casual direction is not as obvious

  • Just showing that an association exists does not solve the dilemma when causal direction is not obvious

  • We often get time order through experimental research studies because, rather than observational, the time order can be fixed or monitored

• Distorted cognitions:

  • people have been blaming strange behavior on a full moon for centuries

• Eliminate alternative explanations

  • we might have not thought of a particular reason that can explain the association

Types of multivariate relationships

• Spurious associations

  • association btwn x and y is spurious if both variables are dependent on a third variable, z, and their association dissapears when z is controlled

  • x and y are associated because both are…

• chain relationships

  • The association between x and y can be understood through a third variable m, mediating, and their association dissapears or reduced when m is controlled

• multiple causes

  • usually multiple variables/outcomes are related to the constant and independent variables

• statistical interactions

  • associations btwn x and y is different within catagories of variaBLE Z WHICH IS A MODERATING VARIABLE

  • moderating variable impact the relationship

  • true effect of x and y changes at different levels of z