Press "Enter" to skip to content

Category: Data Science

Be Careful Of P-Hacking

Published 2018-09-25 by Kevin Feasel

Vincent Granville discusses the problem of p-hacking:

I read an article this morning, about a top Cornell food researcher having 13 studies retracted, see here. It prompted me to write this blog. It is about data science charlatans and unethical researchers in the Academia, destroying the value of p-values again, using a well known trick called p-hacking, to get published in top journals and get grant money or tenure. The issue is widespread, not just in academic circles, and make people question the validity of scientific methods. It fuels the fake “theories” of those who have lost faith in science.

The trick consists of repeating an experiment sufficiently many times, until the conclusions fit with your agenda. Or by being cherry-picking about the data you use, or even discarding observations deemed to have a negative impact on conclusions. Sometimes, causation and correlations are mixed up on purpose, or misleading charts are displayed. Sometimes, the author lacks statistical acumen.

Usually, these experiments are not reproducible. Even top journals sometimes accept these articles, due to

  • Poor peer-review process

  • Incentives to publish sensational material

Wansink is a charlatan.  But beyond p-hacking is Andrew Gelman and Eric Loken’s Garden of Forking Paths.  Gelman’s blog, incidentally (example), is where I originally learned about Wansink’s shady behaviors.  Gelman also warns us not to focus on the procedural, but instead on a deeper problem.

1 Comment

Calculating Lifetime Value With R

Published 2018-09-21 by Kevin Feasel

Sergey Bryl shows how to calculate the lifetime value of a subscription service:

Predicting LTV is a common issue for a new, recently launched product/service/application when we don’t have a lot of historical data but want to calculate LTV as soon as possible. Even though we may have a lot of historical data on customer payments for a product that is active for years, we can’t really trust earlier stats since the churn curve and LTV can differ significantly between new customers and the current ones due to a variety of reasons.

Therefore, regardless of whether our product is new or “old”, we attract new subscribers and want to estimate what revenue they will generate during their lifetimes for business decision-making.

This topic is closely connected to the Cohort Analysis and if you are not familiar with the concept, I recommend that you read about it and look at other articles I wrote earlier on this blog.

Read the whole thing.

Comments closed

Interpreting The Area Under The Receiver Operating Characteristic Curve

Published 2018-09-19 by Kevin Feasel

Roos Colman explains what a Receiver Operating Characteristic (ROC) curve is and how we interpret the Area Under the Curve (AUC):

The AUC can be defined as “The probability that a randomly selected case will have a higher test result than a randomly selected control”. Let’s use this definition to calculate and visualize the estimated AUC.
In the figure below, the cases are presented on the left and the controls on the right.
Since we have only 12 patients, we can easily visualize all 32 possible combinations of one case and one control. (Rcode below)

Expanding from this easy-to-follow example, Colman walks us through some of the statistical tests involved.  Check it out.

Comments closed

Naive Bayes Against Large Data Sets

Published 2018-09-12 by Kevin Feasel

Catherine Bernadorne walks us through using Naive Bayes for sentiment analysis:

The more data that is used to train the classifier, the more accurate it will become over time. So if we continue to train it with actual results in 2017, then what it predicts in 2018 will be more accurate. Also, when Bayes gives a prediction, it will attach a probability. So it may answer the above question as follows: “Based on past data, I predict with 60% confidence that it will rain today.”

So the classifier is either in training mode or predicting mode. It is in training mode when we are teaching it. In this case, we are feeding it the outcome (the category). It is in predicting mode when we are giving it the features, but asking it what the most likely outcome will be.

My contribution is a joke that I heard last night:  a Bayesian statistician hears hooves clomping the ground.  He turns around and sees a tiger.  Therefore, he decides that it must be a zebra.  First time I’d heard that joke, and as a Bayesian zebra-spotter, I enjoyed it.

Comments closed

Disambiguating The Confusion Matrix

Published 2018-09-11 by Kevin Feasel

John Cook walks through a set of valuable terms derived from the core components of the confusion matrix:

How many terms are possible? There are four basic ingredients: TP, FP, TN, and FN. So if each term may or may not be included in a sum in the numerator and denominator, that’s 16 possible numerators and 16 denominators, for a total of 256 possible terms to remember. Some of these are redundant, such as one(a.k.a. ONE), given by TP/TP, FP/FP, etc. If we insist that the numerator and denominator be different, that eliminates 16 possibilities, and we’re down to a more manageable 240 definitions. And if we rule out terms that are the reciprocals of other terms, we’re down to only 120 definitions to memorize.

And of those, John points out the handful which are generally important, providing us an excellent table with definitions of commonly-used terms.

Comments closed

Integrating Kafka Into A Data Scientist’s Workflow

Published 2018-09-07 by Kevin Feasel

Liz Bennett from Stitch Fix has a guest post on the Confluent blog:

Our main requirement for this new project was to build infrastructure that would be 100 percent self-service for our Data Scientists. In other words, my teammates and I would never be directly involved in the discovery, creation, configuration and management of the event data. Self-service would fix the primary shortcoming of our legacy event delivery system: manual administration that was performed by my team whenever a new dataset was born. This manual process hindered the productivity and access to event data for our Data Scientists. Meanwhile, fulfilling the requests of the Data Scientists hindered our own ability to improve the infrastructure. This scenario is exactly what the Data Platform Team strives to avoid. Building self-service tooling is the number one tenet of the Data Platform Team at Stitch Fix, so whatever we built to replace the old event infrastructure needed to be self-service for our Data Scientists. You can learn more about our philosophy in Jeff Magnusson’s post Engineers Shouldn’t Write ETL.

This is an architectural overview and a good read.

Comments closed

Time-Series Analysis With Box-Jenkins

Published 2018-08-30 by Kevin Feasel

The folks at Knoyd walk us through time series analysis using the Box-Jenkins method:

However, this approach is not generally recommended so we have to find something more appropriate. One option could be forecasting with the Box-Jenkins methodology. In this case, we will use the SARIMA (Seasonal Auto Regressive Integrated Moving Average) model. In this model, we have to find optimal values for seven parameters:

  • Auto Regressive Component (p)
  • Integration Component (d)
  • Moving Average Component (q)
  • Seasonal Auto Regressive Component (P)
  • Seasonal Integration Component (D)
  • Seasonal Moving Average Component (Q)
  • Length of Season (s)

To set these parameters properly you need to have knowledge of auto-correlation functions and partial auto-correlation functions.

Read on for a nice overview of this method, as well as the importance of making sure your time series data set is stationary.

Comments closed

Using MLFlow For Binary Classification In Keras

Published 2018-08-29 by Kevin Feasel

Jules Damji walks us through classifying movie reviews as positive or negative reviews, building a neural network via Keras on MLFlow along the way:

François’s code example employs this Keras network architectural choice for binary classification. It comprises of three Dense layers: one hidden layer (16 units), one input layer (16 units), and one output layer (1 unit), as show in the diagram. “A hidden unit is a dimension in the representation space of the layer,” Chollet writes, where 16 is adequate for this problem space; for complex problems, like image classification, we can always bump up the units or add hidden layers to experiment and observe its effect on accuracy and loss metrics (which we shall do in the experiments below).

While the input and hidden layers use relu as an activation function, the final output layer uses sigmoid, to squash its results into probabilities between [0, 1]. Anything closer to 1 suggests positive, while something below 0.5 can indicate negative.

With this recommended baseline architecture, we train our base model and log all the parameters, metrics, and artifacts. This snippet code, from module models_nn.py, creates a stack of dense layers as depicted in the diagram above.

The overall accuracy is pretty good—I ran through a sample of 2K reviews from the set with Naive Bayes last night for a presentation and got 81% accuracy, so the neural network getting 93% isn’t too surprising.  Seeing the confusion matrix in this demo would have been a nice addition.

Comments closed

Dealing With Multicollinearity With R

Published 2018-08-21 by Kevin Feasel

Chaitanya Sagar explains the concept of multicollinearity in linear regressions and how we can mitigate this issue in R:

Perfect multicollinearity occurs when one independent variable is an exact linear combination of other variables. For example, you already have X and Y as independent variables and you add another variable, Z = a*X + b*Y, to the set of independent variables. Now, this new variable, Z, does not add any significant or different value than provided by X or Y. The model can adjust itself to set the parameters that this combination is taken care of while determining the coefficients.

Multicollinearity may arise from several factors. Inclusion or incorrect use of dummy variables in the system may lead to multicollinearity. The other reason could be the usage of derived variables, i.e., one variable is computed from other variables in the system. This is similar to the example we took at the beginning of the article. The other reason could be taking variables which are similar in nature or which provide similar information or the variables which have very high correlation among each other.

Multicollinearity can make regression analysis trickier, and it’s worth knowing about.  H/T R-bloggers.

Comments closed