Post-stratified estimation (stehman 2014)

[adebowaledaniel]

This PR adds a new class for post-stratified estimation of map accuracy metrics, implementing Stehman’s stratified estimators. The class provides methods to compute accuracy metrics (user’s, producer’s, overall, area proportion), their standard errors, and a reporting function for summary output.

New functionality for stratified accuracy estimation:

• Added StehmanStratifiedEstimators class in src/post_stratified_estimation.py to calculate stratified accuracy metrics for map evaluation, including user’s, producer’s, and overall accuracy, as well as area proportion.
• Implemented methods to compute standard errors for all accuracy metrics and F1 score using error propagation, improving statistical rigor in reporting.

Reporting and usability improvements:

• Provided a generate_report method to output a comprehensive DataFrame summary of all metrics and their standard errors for both crop and non-crop classes, supporting dataset and country metadata.

[anatarano]

Looks good to me! I will open an issue so that we can use the standard error to calculate the 95% confidence intervals.