However this is linear in the parameters and not linear in the covariates
as delta_sales_i / delta_discounts_i = 5Discounts^(-0.5) - 5 being the parameter and Discounts being the covariate, therefore any estimator of the such Y = tau(Xi)T + ei should be able to learn that T = 5 and tau = Xi^-0.5
I am not sure that the example in What is Non-Parametric About? is correct
As you stay Sales_i = 20+10*Discounts^0.5 + ei
However this is linear in the parameters and not linear in the covariates
as delta_sales_i / delta_discounts_i = 5Discounts^(-0.5) - 5 being the parameter and Discounts being the covariate, therefore any estimator of the such Y = tau(Xi)T + ei should be able to learn that T = 5 and tau = Xi^-0.5
Am I missing something?