Watch this Ask the Expert session to learn how to use the DEEPPRICE procedure in SAS Econometrics to achieve personalized pricing.
Watch the Webinar
You will learn:
To adapt ideas from machine learning for econometrics in order to answer what-if questions.
How the DEEPPRICE procedure in SAS Econometrics enables you to apply machine learning.
How PROC DEEPPRICE accounts for individual variability among customers for personalized pricing and revenue optimization.
The questions from the Q&A segment held at the end of the webinar are listed below and the slides from the webinar are attached.
Q&A
Does each of the three example value functions represent what DEEPPRICE can accomplish or is that only for one of the three rows?
Value function is a function of policy rule (or treatment). Our aim is to choose the policy rule that maximizes this function. We specify the value function based on the problem that we seek to solve. For example, if treatment is binary, and if the positive treatment effect is preferred, then anyone whose
should be assigned the treatment (see the second row of the table on slide 17). If the outcome is revenue, there is a fixed cost for each treatment, and the final goal is to maximize profit (a portion of the revenue) then, the optimal policy rule should be to assign the treatment if
and not assign it otherwise, where
is the fixed cost of a treatment and
is the profit margin (see the third row of the table on slide 17). In the case study that I covered, the treatment is the price, the outcome is the demand, and the final goal is to maximize the revenue, then the optimal policy rule, i.e., the price, should be set to
(see the fourth row of the table on slide 17).
The first two specifications of the value function do not really apply to DEEPPRICE as they are for a model with a binary treatment variable.
What about the cannibalization effect? Intraction between products? Example: price of product A leads less sales of product B. Are those procs useful for measuring cannibalization?
PROC DEEPPRICE does not take into account any cross-price effects.
Can you please explain the Simpson's paradox slide one more time?
The first plot (the plot on the left on slide 3) is the plot of the amount of exercise individuals had (throughout their lives) against their cholesterol levels. In this plot, you see a positive relationship between exercise and cholesterol, suggesting that more exercise is correlated with more cholesterol. However, if you stratify the observations with respect to age and plot it, as in the plot on the right, then you can clearly see the expected relationship between the amount of exercise and the cholesterol level. Age, here is a confounding factor, and when not accounted for, it can mix up the results so badly that it can even invert the relationship. People often do not recognize the confounders, especially when they work with observational data.
Is instrumental variable supported in DEEPCAUSAL procedure?
Yes. When the treatment variable is endogenous, i.e., there is a confounders or a factor that has impact on both outcome and treatment and is not observed; an instrumental variable should be used.
You can use an instrumental variable for the unobserved confounder in the DEEPPRICE procedure.
Recommended Resources
SAS Econometrics page
The DEEPPRICE Procedure documentation
SAS Econometrics documentation
SAS Econometrics blogs
Please see additional resources in the attached slide deck.
Want more tips? Be sure to subscribe to the Ask the Expert board to receive follow up Q&A, slides and recordings from other SAS Ask the Expert webinars.
... View more
