Statisticians and data scientists want a neat world where data is arranged in a table such that every row is an observation or instance, and every column is a variable or measurement. Getting to this state of “ready to model format” (often called a denormalized form by relational algebra types) often requires quite a bit … Continue reading The case for index-free data manipulation
Imagine that in the course of your analysis, you regularly require summaries of numerical values. For some applications you want the mean of that quantity, plus/minus a standard deviation; for other applications you want the median, and perhaps an interval around the median based on the interquartile range (IQR). In either case, you may want … Continue reading Using replyr::let to Parameterize dplyr Expressions
One thing I teach is: when evaluating the performance of regression models you should not use correlation as your score. This is because correlation tells you if a re-scaling of your result is useful, but you want to know if the result in your hand is in fact useful. For example: the Mars Climate Orbiter … Continue reading Be careful evaluating model predictions
Nina Zumel and I are happy to announce a formal article discussing data preparation and cleaning using the vtreat methodology is now available from arXiv.org as citation arXiv:1611.09477 [stat.AP]. vtreat is an R data.frame processor/conditioner that prepares real-world data for predictive modeling in a statistically sound manner. It prepares variables so that data has fewer … Continue reading vtreat data cleaning and preparation article now available on arXiv
Practical Data Science with R, Zumel, Mount; Manning 2014 is a book Nina Zumel and I are very proud of. I have written before how I think this book stands out and why you should consider studying from it. Please read on for a some additional comments on the intent of different sections of the … Continue reading Teaching Practical Data Science with R
Nina Zumel and I have been doing a lot of writing on the (important) details of re-encoding high cardinality categorical variables for predictive modeling. These are variables that essentially take on string-values (also called levels or factors) and vary through many such levels. Typical examples include zip-codes, vendor IDs, and product codes. In a sort … Continue reading You should re-encode high cardinality categorical variables
We have already written quite a few times about our vtreat open source variable treatment package for R (which implements effects/impact coding, missing value replacement, and novel value replacement; among other important data preparation steps), but we thought we would take some time to describe some of the principles behind the package design. Introduction vtreat … Continue reading Some vtreat design principles
Nina Zumel recently announced upcoming speaking appearances. I want to promote the upcoming sessions at ODSC West 2016 (11:15am-1:00pm on Friday November 4th, or 3:00pm-4:30pm on Saturday November 5th) and invite executives, managers, and other data science consumers to attend. We assume most of the Win-Vector blog audience is made of practitioners (who we hope … Continue reading Data science for executives and managers
Recently Microsoft Data Scientist Bob Horton wrote a very nice article on ROC plots. We expand on this a bit and discuss some of the issues in computing “area under the curve” (AUC). R has a number of ROC/AUC packages; for example ROCR, pROC, and plotROC. But it is instructive to see how ROC plots … Continue reading On calculating AUC
When we teach “R for statistics” to groups of scientists (who tend to be quite well informed in statistics, and just need a bit of help with R) we take the time to re-work some tests of model quality with the appropriate significance tests. We organize the lesson in terms of a larger and more … Continue reading Adding polished significance summaries to papers using R
