Download my free handbook on predictive modeling

I like to keep things simple, so here’s the gist: I wrote another book. It’s free. Download it here.

 

The title says it all: “Cool Data: A how-to guide for predictive modeling for higher education advancement and nonprofits using multiple linear regression in Data Desk.” It’s a 190-page “cookbook,” a guide for folks who aren’t looking for deep understanding of stats, regression, or even predictive modelling, but just enough knowledge — a recipe, really — to mine the value in their organizations’ databases. It’s the kind of book I would have loved to have when I was starting out.

 

Take a look, dive in if it’s your thing, share it with someone who might be interested.

 

I remember talking about the idea as long ago as 2010. I wanted to write something not too technical, yet valid, practical, and actionable. On getting into it I quickly realized that I couldn’t talk about multiple linear regression without talking about how to clean, transform, and prepare data for modelling. And I couldn’t talk about data prep without talking about querying a database. As a result, a large portion of the book is an introduction to SQL; again, not a deep dive into writing queries, but just enough for a motivated person to learn how to build an analysis-ready file.

 

I don’t have to sell you on it, though, because it’s free — download it and do whatever you want with it. If it looks interesting to you, buy the Data Desk software and work through the book using the sample data and your own data. (Be sure to check back for updates to the book which may be necessary as the Data Desk software continues to evolve.) And, of course, consider getting training, preferably one-on-one.

 

Unlike this handbook, Data Desk and training are not free, but they’re investments that will pay themselves back countless times over — if you stick with it.

 

 

Advertisement

Nifty SQL regression to calculate donors’ giving trends

 

Here’s a nifty bit of SQL that calculates a best-fit line through a donor’s years of cash-in giving by fiscal year (ignoring years with no giving), and classifies that donor in terms of how steeply they are “rising” or “falling”.

 

I’ll show you the sample code, which you will obviously have to modify for your own database, and then talk a little bit about how I tested it. (I know this works in Oracle version 11g. Not sure about earlier versions, or other database systems.)

 

with sums AS (
 select t1.id, t1.fiscal_year, log(10, sum(t1.amount)) AS yr_sum
 from gifts t1
 group by t1.id, t1.fiscal_year),

slopes AS (
 select distinct
 sums.id,
 regr_slope(sums.yr_sum,sums.fiscal_year) OVER (partition by sums.id) AS slope

from sums
 )

select
 slopes.id,
 slopes.slope,
 CASE
 when slopes.slope is null then 'Null'
 when slopes.slope >=0.1 then 'Steeply Rising'
 when slopes.slope >=0.05 then 'Moderately Rising'
 when slopes.slope >=0.01 then 'Slightly Rising'
 when slopes.slope >-0.01 then 'Flat'
 when slopes.slope >-0.05 then 'Slightly Falling'
 when slopes.slope >-0.1 then 'Moderately Falling'
 else 'Steeply Falling' end AS description

from slopes

That’s it. Not a lot of SQL, and it runs very quickly (for me). But does it actually tell us anything?

 

I devised a simple test. Adapting this query, I calculated the “slope of giving” for all donors over a five-year period in the past: FY 2007 to FY 2011. I wanted to see if this slope could predict whether, and by how much, a donor’s giving would rise or fall in the next five-year period: FY 2012 to FY 2016. (Note that the sum of a donor’s giving in each year is log-transformed, in order to better handle outlier donors with very large giving totals.)

 

I assembled a data file with each donor’s sum of cash giving for the first five-year period, the slope of their giving in that period, and the sum of their cash giving for the five-year period after that.

 

The first test was to see how the categories of slope, from Steeply Rising to Steeply Falling, translated into subsequent rises and falls. In Data Desk, I compared the two five-year periods. If the second period’s giving was greater than the first, I called that a “rise.” If it was less, I called it a “fall.” And if it was exactly the same, I called it “Same.”

 

The table below summarizes the results. Note that these numbers are all percentages, summed horizontally. (I will explain the colour highlighting later on.)

 

 

For Steeply Rising, 60.6% of donors actually FELL from the first period to the next. Only 37.8 percent rose, and just 1.6% stayed exactly the same. Not terribly impressive. Look at Steeply Falling, though: More than three-quarters actually did fall. That’s a better result, but then again, “Falling” dominates for every category; in the whole file, close to 70% of all donors reduced their giving in the next period. If a donor has no giving in the second period of five years, that’s zero dollars given, and this is called a “Fall” — more on that aspect in just a sec.

 

(I’ve left out donors with a FY2007-11 slope of Null — they’re the ones who gave in only one year and therefore don’t have a “slope”.)

 

Let’s not give up just yet, however. The colour highlighting indicates how high each percentage value is in relation to those above and below it. For example, the highest percentages in the Falling column are found in the Slightly, Moderately, and especially Steeply Falling slope categories. The highest percentages in the Rising column are in the Slightly, Moderately, and Steeply Rising slope categories. And in the Same column, the Flat slope wins hands-down — as we would hope.

 

So a rising slope “sort of” predicts increased giving, a falling slope “sort of” predicts decreased giving. Unfortunately, many donors are not retained into the second five-year period, so there’s not a lot to be confident about.

 

But what if a donor IS retained? What if we exclude the lapsed donors entirely? Let’s do that:

 

 

Excluding non-donors seems to lead to an improvement … The slope does a better job sorting between the risers and fallers when a donor is actually retained. Again, the colour highlighting is referencing columns, not rows. But notice now that, across the rows, Rising has a slight majority for the Rising slope categories, and Falling has a slight majority for the Falling slope categories. (The bar is set too high for Flat, however, given that a donor’s giving in the first five years has to be exactly equal to her giving in the second five years to be called Same.)

 

Admittedly, these majorities are not generous. If I calculated a donor’s slope of giving as Steeply Rising and that donor was retained, I have only a 56.4% chance of actually being right. And of course there’s no guarantee that donor won’t lapse.

 

(Note that these are donors of all types — alumni, non-alumni individuals, and entities such as corporations and foundations. Non-alumni donors tend not to have patterns in their giving that are repeated, not to the extent that alumni do. However, when I limit the data file to alumni donors only, the improvement in this method is only very slight.)

 

Pressing on … I did a regression analysis using total giving in the second five-year period as the dependent variable, then entered total giving in the prior five-year period as an independent variable. (Naturally, R-squared was very high.) This allowed me to see if Slope provides any explanatory power when it is added as the second independent variable — the effect of giving in the first five-year period already being accounted for.

 

And the answer is, yes, it does. But only under specific conditions: Both five-year giving totals were log-transformed and, most significantly, donors who did not give in the second period were excluded from the regression.

 

There are other way to assess the usefulness of “slope” which might lead to an application, and I encourage you to give this a try with your own data. From past experience I know that donors who make big upgrades in giving don’t have any neat universal pattern such as an upward slope in their giving history. (The concept of volatility is explored here and here.) “Slope” is probably too simple a characteristic to employ on its own.

 

But as I’ve said before, if it were easy, obvious, or intuitive, it wouldn’t be data analysis.

 

CoolData (the book) beta testers needed

 

UPDATE (Jan 5): 16 people have responded to my call for volunteers, so I am going to close this off now. I have been in touch with each person who has emailed me, and I will be making a final selection within a few days. Thank you to everyone who considered taking a crack at it.

 

Interested in being a guinea pig for my new handbook on predictive modelling? I’m looking for someone (two or three people, max) to read and work through the draft of “CoolData” (the book), to help me make it better.

 

What’s it about? This long subtitle says it all: “A how-to guide for predictive modelling for higher education advancement and nonprofits using multiple linear regression in Data Desk.”

 

The ideal beta tester is someone who:

 

• has read or heard about predictive modelling and understands what it’s for, but has never done it and is keen to learn. (Statistical concepts are introduced only when and if they are needed – no prior stats knowledge is required. I’m looking for beginners, but beginners who aren’t afraid of a challenge.);
• tends to learn independently, particularly using books and manuals to work through examples, either in addition to training or completely on one’s own;
• does not have an IT background but has some IT support at his or her organization, and would not be afraid to learn a little SQL in order to query a database him- or herself, and
• has a copy of Data Desk, or intends to purchase Data Desk. (Available for PC or Mac).

 

It’s not terribly important that you work in the higher ed or nonprofit world — any type of data will do — but the book is strictly about multiple linear regression and the stats software Data Desk. The methods outlined in the book can be extended to any software package (multiple linear regression is the same everywhere), but because the prescribed steps refer specifically to Data Desk, I need someone to actually go through the motions in that specific package.

 

Think of a cookbook full of recipes, and how each must be tested in real kitchens before the book can go to press. Are all the needed ingredients listed? Has the method been clearly described? Are there steps that don’t make sense? I want to know where a reader is likely to get lost so that I can fix those sections. In other words, this is about more than just zapping typos.

 

I might be asking a lot. You or your organization will be expected to invest some money (for the software, sales of which I do not benefit from, by the way) and your time (in working through some 200 pages).

 

As a return on your investment, however, you should expect to learn how to build a predictive model. You will receive a printed copy of the current draft (electronic versions are not available yet), along with a sample data file to work through the exercises. You will also receive a free copy of the final published version, with an acknowledgement of your work.

 

One unusual aspect of the book is that a large chunk of it is devoted to learning how to extract data from a database (using SQL), as well as cleaning it and preparing the data for analysis. This is in recognition of the fact that data preparation accounts for the majority of time spent on any analysis project. It is not mandatory that you learn to write queries in SQL yourself, but simply knowing which aspects of data preparation can be dealt with at the database query level can speed your work considerably. I’ve tried to keep the sections about data extraction as non-technical as possible, and augmented with clear examples.

 

For a sense of the flavour of the book, I suggest you read these excerpts carefully: Exploring associations between variables and Testing associations between two categorical variables.

 

Contact me at kevin.macdonell@gmail.com and tell me why you’re interested in taking part.

 

 

 

Targeting rare behavior

Guest post by Kelly Heinrich, Assistant Director of Prospect Management and Analytics, Stanford University

Last August, about two months into a data analyst position with a university’s development division, I had the task to build a predictive model for the Office of Gift Planning (OGP). The OGP wanted a tool to help them focus on the constituents who are most likely to make a planned gift. I wanted to identify a few hundred of the best planned giving prospects who I could prioritize by the probability of donating. After a bit of preliminary research, I chose: 1) 50 years of age and older and 2) inclusion in a recent wealth screening as the criteria for the study population. This generated a file of 133,000 records; 582 of them were planned gift donors. I’ve worked with files larger than this and did not expect a problem. However, that turned out to be a mistake because the planned gift donors, who exhibited the target behavior, comprised 0.4% of the population, a proportion so small it can be considered rare. I’ll explain more about that later; first I want to describe the project as it developed.

I decided to use logistic regression with the dependent variable being either “made a planned gift” or “has not made a planned gift”. I cleaned the data and identified some strong relationships between the variables. After trying several combinations for the regression model, I had one with a Nagelkerke of .24, which is relatively good. (Nagelkerke is like a pseudo R squared; it can be loosely interpreted as the variability of the dependent variable that is accounted for by the model’s independent variables.) However, when I applied the algorithm to the study population, only 31 constituents without a planned gift and only 11 planned giving donors were identified as having a probability of giving of .5 or greater. I lowered the probability threshold of giving to .2 or greater and 105 non-planned givers and 52 planned gift donors fell into this range. This was still disappointing.

Desperate to identify more new potential prospects, I explored more criteria to narrow the study population and built three successive models. For the purpose of the follow-up exploratory research and this article, I re-built all four models using the same independent variables to easily compare their outcomes. Here’s a summary of the four models:

Models B, C, and D are all subsets of the original data set. Each model has advantages and disadvantages to it and I was uncertain how to evaluate them against one another. For example, each additional filtering criterion resulted in losing part of the target population, meaning that I systematically eliminated constituents with characteristics that are in fact associated with making a planned gift. I scored everyone who was identified with a probability of .2 or greater in any of the models by the number of models in which they were identified. I’m not unhappy with that solution, but since then I’ve been learning about better methods for targeting rare behavior.

If the OGP was interested only in prioritizing the prospects already in their pool of potential planned giving donors, model D would serve their need. However, we wanted to identify the best potential planned giving prospects within the database. If we want to uncover untapped potential in an ever-growing database, we need to explore methods on how to target rare behavior. This seems especially important in our field where 1) donating, in general, is somewhat rare and 2) donating really generous gifts is rarer. Better methods of targeting rare behavior will also be useful for modeling for special initiatives and unique kinds of gifts.

As I’ve been learning, logistic regression suffers from small sample bias when the target behavior is rare, relative to the study population. This helps explain why applying the algorithm to the original population resulted in very few new prospects–even though the model had a decent Nagelkerke of .24. Some analysts suggest using alternative sampling methods when the target behavior comprises less than 5% of the study. (See endnote.) Knowing that the planned gift donors in my original project comprised only 0.4% of the population, I decided to experiment with two new approaches.

In both of the exploratory models, I created the study population size so planned gift donors would comprise 5 percent. First, I generated a study population by including all 582 of the planned gift donors and a random selection of 11,060 non-planned-gift constituents (model E). Then, I applied the algorithm from that population to the entire non-planned-gift population of 132,418. In the second approach (model F), the planned gift population was randomly split into two equal size groups of 291. I also randomly selected 5,530 non-planned-gift constituents. To build the regression model, I combined one of the planned gift donor groups (of 291) with 5,530 non-planned-gift constituents. I then tested the algorithm on the holdout sample (the other planned giving group of 291 with 5,530 non-planned-gift constituents). Finally, I applied the algorithm to the entire original population of 133,000. Here are the results:

Using the same independent variables as in models A through D, model E had a Nagelkerke of .39 and model F .38, which helps substantiate that the independent variables are useful predictors for planned giving. Models E and F were more effective at predicting the planned givers (129 and 123 respectively with a probability of giving greater than or equal to .5) compared to model A (11), i.e. more than ten times as many. The sampling techniques have some advantages and disadvantages. The disadvantage is that by reducing the non-planned-gift population, it loses some of its variability and complexity. However, the advantage, in both models E and F, is that 1) the target population maintains its complexity, 2) new prospects are not limited by characteristic selection (the additional criteria that I used to reduce the population in models B, C, and D), which increases the likelihood of identifying constituents who were previously not on the OGP’s radar, and 3) the effects of the sample bias seem to be reduced.

It’s important to note that I displayed the measures (Nagelkerke and estimated probabilities) from the exploratory models and populations purely for comparison purposes. Because the study population is manipulated in the exploratory methods, the probability of giving should not be directly interpreted as actual probabilities. However, they can be used to prioritize those with the highest probabilities and that will serve our need.

To explore another comparison between models A and F, I ranked all 133,000 records in each. I then sorted all the records in model F in descending order. I took the top 1,000 records from model F and then ran correlation between the rank of model A and the rank of model F; they have a correlation of .282, meaning there is a substantial difference between the ranked records.

Over the last several months, Peter Wylie, Higher Education Consultant and Contractor, and I have been exchanging ideas on this topic. I thank him for his insight, suggestions, and encouragement to share my findings with our colleagues.

It would be helpful to learn about the methods you’ve used to target rare behavior. We could feel more confident about using alternative methods if repeat efforts produced similar outcomes. Furthermore, I did not have a chance to evaluate the prospecting performance of these models, so if you have used a method for targeting rare behavior and have had an opportunity to assess its effectiveness, I am very interested in learning about that. I welcome ideas, feedback, examples from your research, and questions in regard to this work. Please feel free to contact me at heinrichkellyl@gmail.com.

Endnotes

The ideas for these alternative approaches are adapted from the following articles:

• Chawla, Nitesh, Aleksandar Lazarevic, Lawrence Hall, and Kevin Bowyer. 2003. “SMOTEBoost: Improving Prediction of the Minority Class in Boosting.” http://www3.nd.edu/~dial/papers/ECML03.pdf
• King, Gary and Lanche Zeng. 2001. “Logistic Regression in Rare Events Data.” http://gking.harvard.edu/files/abs/0s-abs.shtml

Kelly Heinrich has been conducting quantitative research and analysis in higher education development for two and a half years. She has recently accepted a position as Assistant Director of Prospect Management and Analytics with Stanford University that will begin in June 2013.

Logistic vs. multiple regression: Our response to comments

Guest post by John Sammis and Peter B. Wylie

Thanks to all of you who read and commented on our recent paper comparing logistic regression with multiple regression. We were not sure how popular this topic would be, but Kevin told us that interest was high, and there were a number of comments and questions. There were several general themes in the comments; Kevin has done an excellent job responding, but we thought we should throw in our two cents.

Why not just use logistic?

The point of our paper was not to suggest that logistic regression should not be used — our point was that multiple regression can achieve prediction results quite similar to logistic regression. Based on our experience working with and training fundraising professionals getting introduced to analytics, logistic regression can be intimidating. Our goal is always to get these folks to use analytics to help with their fundraising initiatives. We find many of them catch on with multiple regression, and much less so with logistic regression.

Predicted values vs. probabilities

We understand that the predicted values generated by multiple regression are different from the probabilities generated by logistic regression. Regardless of the statistic modeling technique we use, we always bin the raw prediction or probability values into equal-sized score levels. We have found that score level bins are easier to use than raw values. And using equal-sized score levels allows for easier evaluation of the scoring model.

“I cannot agree”

Some commenters, knowledgeable about statistics, said they would not use multiple regression when the inputs called for logistic. According to the rules, if the target variable is binary, then linear modelling doesn’t make sense — and the rules must be obeyed. In our view, this rigid approach to method selection is inappropriate for predictive modelling. The use of multiple linear regression in place of logistic regression may not always make theoretical sense, but predictive modellers are concerned with whether or not a model produces an output that is useful in practical terms. The worth of a model is testable against new, real-world data, therefore a model has only one criterion for determining “appropriate” use: Whether it really predicts what the modeler claims it will predict. The truth is revealed during evaluation.

A modest proposal

No one reading this should simply take our word that these two dissimilar methods yield similar results. Neither should anyone dismiss it out of hand without providing a critique based on real data. We would encourage anyone to try doing something on your own with data using both techniques and show us what you find. In particular, graduate students looking for a thesis or dissertation topic might consider producing something under this title: “Comparing Logistic Regression and Multiple Regression as Techniques for Predicting Major Giving.”

Heck! Peter says that if anyone were interested in doing a study like this for a thesis or dissertation, he would be willing to offer advice on how to:

1. Do a thorough literature review
2. Formulate specific research questions
3. Come up with a study design
4. Prepare a proposal that would satisfy a thesis or dissertation committee.

That’s quite an offer. How about it?

When less data is more, in predictive modelling

When I started doing predictive modelling, I was keenly interested in picking the best and coolest predictor variables. As my understanding deepened, I turned my attention to how to define the dependent variable in order to really get at what I was trying to predict. More recently, however, I’ve been thinking about refining or limiting the population of constituents to be scored, and how that can help the model.

What difference does it make who gets a propensity score? Up until maybe a year ago, I wasn’t too concerned. Sure, probably no 22-year-old graduate had ever entered a planned giving agreement, but I didn’t see any harm in applying a score to all our alumni, even our youngest.

Lately, I’m not so sure. Using the example of a planned gift propensity model, the problem is this: Young alumni don’t just get a score; they also influence how the model is trained. If all your current expectancies were at least 50 before they decided to make a bequest, and half your alumni are under 30 years old, then one of the major distinctions your model will make is based on age. ANY alum over 50 is going to score well, regardless of whether he or she has any affinity to the institution, simply because 100% of your target is in that age group.

The model is doing the right thing by giving higher scores to older alumni. If ages in the sample range from 21 to 100+, then age as a variable will undoubtedly contribute to a large chunk of the model’s ability to “explain” the target. But this hardly tells us anything we didn’t already know. We KNOW that alumni don’t make bequest arrangements at age 22, so why include them in the model?

It’s not just the fact that their having a score is irrelevant. I’m concerned about allowing good predictor variables to interact with ‘Age’ in a way that compromises their effectiveness. Variables are being moderated by ‘Age’, without the benefit of improving the model in a way that we get what we want out of it.

Note that we don’t have to explicitly enter ‘Age’ as a variable in the model for young alumni to influence the outcome in undesirable ways. Here’s an example, using event attendance as a predictor:

Let’s say a lot of very young alumni and some very elderly constituents attend their class reunions. The older alumni who attend reunions are probably more likely than their non-attending classmates to enter into planned giving agreements — for my institution, that is definitely the case. On the other hand, the young alumni who attend reunions are probably no more or less likely than their non-attending peers to consider planned giving — no one that age is a serious prospect. What happens to ‘event attendance’ as a predictor in which the dependent variable is ‘Current planned giving expectancy’? … Because a lot of young alumni who are not members of the target variable attended events, the attribute of being an event attendee will be associated with NOT being a planned giving expectancy. Or at the very least, it will considerably dilute the positive association between predictor and target found among older alumni.

I confirmed this recently using some partly made-up data. The data file started out as real alumni data and included age, a flag for who is a current expectancy, and a flag for ‘event attendee’. I massaged it a bit by artificially bumping up the number of alumni under the age of 50 who were coded as having attended an event, to create a scenario in which an institution’s events are equally popular with young and old alike. In a simple regression model with the entire alumni file included in the sample, ‘event attendance’ was weakly associated with being a planned giving expectancy. When I limited the sample to alumni 50 years of age and older, however, the R squared statistic doubled. (That is, event attendance was about twice as effective at explaining the target.) Conversely, when I limited the sample to under-50s, R squared was nearly zero.

True, I had to tamper with the data in order to get this result. But even had I not, there would still have been many under-50 event attendees, and their presence in the file would still have reduced the observed correlation between event attendance and planned giving propensity, to no useful end.

You probably already know that it’s best not to lump deceased constituents in with living ones, or non-alumni along with alumni, or corporations and foundations along with persons. They are completely distinct entities. But depending on what you’re trying to predict, your population can fruitfully be split along other, more subtle distinctions. Here are a few:

• For donor acquisition models, in which the target value is “newly-acquired donor”, exclude all renewed donors. You strictly want to have only newly-acquired donors and never-donors in your model. Your good prospects for conversion are the never-donors who most resemble the newly-acquired donors. Renewed donors don’t serve any purpose in such a model and will muddy the waters considerably.
• Conversely, remove never-donors from models that predict major giving and leadership-level annual giving. Those higher-level donors tend not to emerge out of thin air: They have giving histories.
• Looking at ‘Age’ again … making distinctions based on age applies to major-gift propensity models just as it does to planned giving propensity: Very young people do not make large gifts. Look at your data to find out at what age donors were when they first gave $1,000, say. This will help inform what your cutoff should be.
• When building models specifically for Phonathon, whether donor-acquisition or contact likelihood, remove constituents who are coded Do Not Call or who do not have a valid phone number in the database, or who are unlikely to be called (international alumni, perhaps).
• Exclude international alumni from event attendance or volunteering likelihood models, if you never offer involvement opportunities outside your own country or continent.

Those are just examples. As for general principles, I think both of the following conditions must be met in order for you to gain from excluding a group of constituents from your model. By a “group” I mean any collection of individuals who share a certain trait. Choose to exclude IF:

1. Nearly 100% of constituents with the trait fall outside the target behaviour (that is, the behaviour you are trying to predict); AND,
2. Having a score for people with that trait is irrelevant (that is, their scores will not result in any action being taken with them, even if a score is very low or very high).

You would apply the “rules” like this … You’re building a model to predict who is most likely to answer the phone, for use by Phonathon, and you’re wondering what to do with a bunch of alumni who are coded Do Not Call. Well, it stands to reason that 1) people with this trait will have little or no phone contact history in the database (the target behaviour), and 2) people with this trait won’t be called, even if they have a very high contact-likelihood score. The verdict is “exclude.”

It’s not often you’ll hear me say that less (data) is more. Fewer cases in your data file will in fact tend to depress your model’s R squared. But your ultimate goal is not to maximize R squared — it’s to produce a model that does what you want. Fitting the data is a good thing, but only when you have the right data.