Interpreting mean estimate for continous treatment

[mrshanth]

I have a treatment variable, with values [100,500,1000]

The mean estimate for continuous is 0.5,

The average of outcome for treatment cohort is 540, and for control cohort is 90.

Is the interpretation of mean estimate 0.5, then the true beta of treatment on outcome? Estimate method used was linear_regression

[amit-sharma]

If you provided the treatment as a continuous variable, then it means that each unit increase in treatment leads to a 0.5 (half-unit) increase in outcome. So going from 100 to 500 (change of 400) will increase outcome by 200. So yes, you can interpret it as the true beta under linear regression (assuming that the DGP was linear and your graph is correct)

However, if your treatment has only three values, it may also make sense to consider it as a discrete variable.