Causal Assumptions Check
The contents of the Causal Assumptions Check section in the Causal Treatment model report depends on the treatment type.
Binary Treatment
When the treatment is binary, the Causal Assumptions Check section contains three subsections: Positivity, Covariate Balance, and Absolute Standardized Mean Difference Table.
Positivity
The Positivity density plot shows propensity score on the horizontal axis and percentage on the vertical axis. The blue curve shows the propensity score distribution of the control group and the red curve shows the propensity score distribution of the treatment group. The propensity score is the estimated probability of receiving treatment. To interpret the average treatment effect estimate causally, the positivity assumption must be met. The positivity assumption says that every observation in the population has a nonzero probability of receiving each level of treatment. If either curve shows observations congregating near 0 or 1, the positivity assumption might be violated. If the Show Tips and Interpretations option in the Fit Causal Treatment red triangle menu is selected, a more detailed explanation of the positivity assumption appears under the plot.
Covariate Balance
The Covariate Balance plot shows Absolute Standardized Mean Difference (ASMD) on the horizontal axis and variable on the vertical axis. This plot shows variation between the two treatment levels with regard to each variable. The ASMD is the variation measure. It is calculated for each variable by taking the absolute difference in means between the treated and untreated groups and dividing by the standard deviation of the treated group.
If the treatment model is specified, the variables on the vertical axis correspond to the variables included in the treatment model. The red squares show the ASMD values before inverse probability weighting with ratio adjustment (IPWR). The blue dots show the IPW Adjusted ASMD values. The blue dots should be closer to 0 than the red squares, since IPW adjustment is meant to balance the treatment and control groups on all variables besides the treatment, as if the treatment were randomly assigned.
If only the response model is specified, the variables on the vertical axis correspond to the variables included in the response model (excluding the treatment). Since the treatment model is not specified, no IPW adjustment is performed. Therefore, the only points on this plot are blue dots, which show the ASMD between the two treatment levels for each variable.
If the Show Tips and Interpretations option in the Fit Causal Treatment red triangle menu is selected, a description of ASMD appears below the plot.
Absolute Standardized Mean Difference Table
If the treatment model is specified, the Absolute Standardized Mean Difference Table shows the Original and IPW Adjusted ASMD between the two treatment levels for each variable in the treatment model. If only the response model is specified, the table shows only the Original ASMD between the two treatment levels for each variable in the response model.
Continuous Treatment
When the treatment is continuous, the Causal Assumptions Check section contains a Positivity ridgeline plot. The Positivity ridgeline plot shows the GPS distributions at each specified treatment level. The treatment levels on the vertical axis correspond to the Treatment Level Minimum, Treatment Level Maximum, and Treatment Level Increment values in the Model Information section. If the Show Tips and Interpretations option is checked in the Fit Causal Treatment red triangle menu, an interpretation in context of the positivity assumption appears below the plot.