Why is Regularization Underused? An Empirical Study on Trust and Adoption of Statistical Methods

Abstract

Statistical practice does not automatically follow methodological innovation. Regularization methods, widely advocated to reduce overfitting and stabilize inference, are readily available in modern software, but are not consistently used by data analysts. We investigate this implementation gap in a large-scale empirical study of trust in, and acceptance of, regularization techniques, based on $N = 606$ data analysts. Drawing on measurement frameworks from technology acceptance research, we survey practitioners and embed a randomized experiment to test whether written recommendation of regularization methods increases trust or intended use. We find no evidence of such an effect. Instead, adoption intentions are strongly associated with analysts' perceptions of ease of implementation and practical benefit, such as improved bias control or interpretability. Perceived social norms also emerge as a central driver. These results indicate that uptake of statistical methodology depends less on formal recommendations than on usability, perceived utility, and community practice.

Figures (7)

• Figure 1: Distribution of the 5-point Likert items in the order given in Table \ref{['tab:items']}. The percentage on the left hand side/middle/right hand side shows the (combined) share of response categories 1 and 2/3/4 and 5.
• Figure 2: Empirical Kendall's $\tau$ correlation matrix of the investigated constructs. Constructs are sorted in descending order of their correlation with bi.
• Figure 3: Joint distribution of demographic variables with bi.
• Figure 4: Scores for the constructs trust (tr), vigilance (vi), and behavioural intention (bi) per recommendation groups Control, Expert, Journal, and Peer.
• Figure 5: 10-fold cross validation Brier score in a cumulative logit model with bi as the outcome variable and the following predictors (main effects only): gender, age, ee, su, pe, at, ex, tr, si, and vi. Grey bars: standard error of the respective cross validation average. Vertical dashed orange line: optimum. Vertical dashed blue line: largest $\lambda$, where the score is within one standard error of the optimum.