By Gabriel Vasconcelos
In this post I am going to discuss some features of Regression Trees an Random Forests. Regression Trees are know to be very unstable, in other words, a small change in your data may drastically change your model. The Random Forest uses this instability as an advantage through bagging (you can see details about bagging here) resulting on a very stable model.
We are preparing an R for Data-Science course (direct link here) in partnership with the IBPAD (Brazilian Institute of Research and Data Analysis). It is a great course for those who want to have a solid start in R. No prior knowledge in statistics, calculus or programming is required.
In many cases, especially if you are dealing with forecasting models, it is natural to use a large set of models to forecast the same variable and then select the best model using some error measure. For example, you can break the sample into a training sample (in-sample) and a test sample (out-of-sample), estimate all models in the training sample and see how the perform in the test sample. You could compare the models using the root mean squared error (RMSE) or the mean absolute error (MAE).
By Yuri Fonseca
In this post we will discuss briefly about pricing optimization. The main idea behind this problem is the following question: As manager of a company/store, how much should I charge in order to maximize my revenue or profit?
By Gabriel Vasconcelos
When I first started using R there was no such thing as the tidyverse. Although some of the tidyverse packages were available independently, I learned to treat my data mostly using brute force combining pieces of information I had from several sources. It is very interesting to compare this old school programming with the tidyverse writing using the magrittr package. Even if you want to stay old school, tidyverse is here to stay and it is the first tool taught in many data science courses based on R.
By Gabriel Vasconcelos
In a late post I discussed the Double Selection (DS), a procedure for inference after selecting controls. I showed an example of the consequences of ignoring the variable selection step discussed in an article by Belloni, Chernozhukov and Hansen.
By Gabriel Vasconcelos
In a recent paper, which can be downloaded here, Carvalho, Masini and Medeiros show that estimating counterfactuals in a non-stationary framework (when I say non-stationary it means integrated) is a tricky task. It is intuitive that the models will not work properly in the absence of cointegration (spurious case), but what the authors show is that even with cointegration, the average treatment effect (ATE) converges to a non-standard distribution. As a result, standard tests on the ATE will identify treatment effects in several cases when there is no effect at all.
R-Bloggers Feed
