|
return pd.Series(scipy.stats.norm.cdf(s), index=s.index) |
Current Code:
return pd.Series(scipy.stats.norm.cdf(s), index=s.index)
Suggested Replacement:
from scipy.special import ndtr
return pd.Series(ndtr(s), index=s.index)
The current implementation uses scipy.stats.norm.cdf(s) to compute the cumulative distribution function (CDF) of a standard normal distribution. While this approach is correct and expressive, it introduces unnecessary overhead because norm.cdf constructs and evaluates a frozen distribution object under the hood.
In contrast, scipy.special.ndtr(s) provides a low-level, highly optimized C implementation that computes the standard normal CDF directly. It avoids object creation and internal delegation, making it significantly faster—especially when applied to large arrays or inside performance-critical loops.
Since both norm.cdf(s) and ndtr(s) produce numerically identical results for the standard normal distribution, replacing one with the other is a safe and efficient optimization.
pyjanitor/janitor/math.py
Line 229 in c1a96e5
Current Code:
return pd.Series(scipy.stats.norm.cdf(s), index=s.index)Suggested Replacement:
The current implementation uses scipy.stats.norm.cdf(s) to compute the cumulative distribution function (CDF) of a standard normal distribution. While this approach is correct and expressive, it introduces unnecessary overhead because norm.cdf constructs and evaluates a frozen distribution object under the hood.
In contrast, scipy.special.ndtr(s) provides a low-level, highly optimized C implementation that computes the standard normal CDF directly. It avoids object creation and internal delegation, making it significantly faster—especially when applied to large arrays or inside performance-critical loops.
Since both norm.cdf(s) and ndtr(s) produce numerically identical results for the standard normal distribution, replacing one with the other is a safe and efficient optimization.