For background reading on the value of finding “positive deviants” see these resources:
- The Positive Deviance Initiative website
- Wikipedia on Positive Deviance
- “The Power of Positive Deviance: How Unlikely Innovators Solve the World’s Toughest Problems“
This brief post outlines how EvalC3 can help find cases which may be usable examples of Positive Deviance.
- First, develop a predictive model that is good at predicting the absence of the outcome of interest. Usually, we are trying to predict its presence. This can be done by using EvalC3’s search algorithms. Or it can be done by testing out combinations of attributes that according to our prior knowledge and theory are conducive to the outcome not occurring – especially attributes of this kind that we think are quite prevalent.
- Then focus on the False Positives i.e. those cases where the model attributes predicted the absence of the outcome but in practice the outcome was present. These cases qualify, on first glance, as Positive Deviants. They are the cases where it would be well worthwhile doing a within-case investigation in order to find out how they managed to succeed against the odds.
- Try to minimise, but not totally eliminate, the number of False Positives. If there are a lot of False Positives all this may tell us is that the current prediction model is not very good, and is lacking some important attributes. If there are very few, perhaps only one, it is more likely this is a genuine Positive Deviance case achieving the outcome despite all the odds being against it doing so
- Try to minimise the number of False Negatives. This is not essential, but the wider the coverage of the prediction model the more likely the Positive Deviance case will be of wider interest
- Carry out a within-case investigation of the identified Positive Deviance case, (a) to verify if it has been accurately described and thus correctly classified as False Positive, (b) to identify any causal mechanism at work that can explain its performance.
This approach can be tested out using the Krook data set , which is built into EvalC3. The absence of quotas for women in parliament is sufficient for low levels of women’s participation in parliament. It predicts 13 of the 14 countries with such low levels. The one exception is Lesotho, where there are no quotas but there are high levels of participation of women in parliament. This is an example of a “positive deviance” case that would be worth doing within-case investigations to identify and understand the causal processes at work.
Outliers of different kinds
Positive deviance cases are one kind of outlier. But there different kinds of outliers can be found in the contents of a Confusion Matrix. At one level there are the False Positive and False Negative cases, if they are in a minority compared to numbers of True Positives and True Negatives respectively. At another more detailed level, within each of the four Confusion Matrix categories, we can find both modal and outlier cases. To find examples of this latter category of outliers go to the View Cases view and click on Calculate Similarity and then look for cases that have the lowest Similarity measure within their Confusion Matrix category. These are worth investigating as part of a case comparison strategy discussed here, as the end point of an EvalC3 analysis.
Postscript 2018 04 18: Here is a paper that has been waiting to be published, and is well worth reading…”Searching for Success: A Mixed Methods Approach to Identifying and Examining Positive Outliers in Development Outcomes” by Caryn Peiffer and Rosita Armytage, April 2018. Well worth reading, on how and why a combination of quantitative and qualitative analysis is the best way to identify positive outliers (aka positive deviants) and the reasons why some of these might not otherwise see the light of day.