Categories → Causal Wizard Concept , Graph , Causal Inference , Variables , Study Design
In causal inference, treatment refers to the intervention or manipulation that is being studied, with the aim of determining its causal effect on an outcome of interest.
In Causal Wizard, you must specify two variables in your data:
Usually, you will be wanting to estimate the effect of the Treatment on the Outcome. The effect may be partial, or full; it may be direct, or indirect.
Treatment can take many forms, such as a medication, a behavioral intervention, a policy change, or any other kind of intervention.
The treatment variable is the variable that represents the presence or absence of the treatment in a study. It is usually a binary variable that takes the value 1 or True if the participant received the treatment and 0 or False if they did not. In some cases, the treatment variable can also take on other values, such as different doses or types of treatment (categorical). In causal Wizard, the app will try to establish either specific values or a threshold to divide your samples into Control (did not receive treatment) and Treated groups.
In some literature, the Treatment may also be described as an Exposure (usually if the Treatment occurred incidentally) or Intervention (if the Treatment was purposeful).
The outcome variable is the one which is (possibly) affected by the treatment variable. You're using Causal Wizard to try to measure the effect of Treatment on Outcome. Therefore, the outcome variable must not be the same as the treatment variable.
You can use the value of a variable at time A as the treatment, and the value of the same variable at time B as the Outcome, as long as they are represented in your data as different columns (i.e. will be modelled as different variables). A specific method, known as the Difference-in-Differences method, uses this technique to estimate causal effects.
In experimental design, the treatment would be the independent variable, and the outcome would be the dependent variable.
A simple example: A study that investigates the effect of a new medication on reducing blood pressure. In this study, the treatment is the medication, and the treatment variable is a binary variable that indicates whether the participant received the medication or a placebo. Participants who received the medication are coded as 1, and those who received the placebo are coded as 0. The outcome variable is the reduction in blood pressure, and the aim of the study is to determine the causal effect of the medication on this outcome.
In Causal Wizard, your domain knowledge is used to identify the treatment variable.