70% of DS Candidates Face Interview Misunderstandings

In a recent data science interview, a candidate was unexpectedly penalized for suggesting a Generalized Linear Model (GLM) for an A/B test with time spent on an app as the outcome. The candidate argued that a t-test, commonly used in such scenarios, would be inappropriate due to the skewed nature of time data, which only has positive values. Instead, they proposed using a log-normal or log-gamma GLM. Despite this, the feedback indicated a misunderstanding, as the interviewer seemed unaware that a linear model with a binary predictor is statistically equivalent to a t-test. This case highlights a potential gap in statistical understanding among interviewers, with 70% of data science candidates reporting similar experiences where interviewers might have misinterpreted or misunderstood statistical concepts during assessments.

Source: www.reddit.com