ilyesrezgui/AirQo-African-Air-Quality

Estimation and validation of PM2.5 levels using satelitte and ground observations in selected african cities for env justice # AirQo-African-Air-Quality (2023/2024 Data) Estimation and validation of PM2.5 levels using satelitte and ground observations in selected african cities for env justice Data Analysis: 1- Skewness of the PM2.5 variable in the training data: We can see that the PM2.5 has a right skew. * Original pm2_5: Skewness = 7.383403684525106, Kurtosis = 86.1193415970843 * Log Transformed: Skewness = 0.49138261989345633, Kurtosis = 1.4359962407868387 * Square Root Transformed: Skewness = 2.529693698583593, Kurtosis = 14.19940623948297 * Box-Cox Transformed: Skewness = -0.03428424477994078, Kurtosis = 0.9151702451491679 * Yeo-Johnson Transformed: Skewness = -0.03428420696278356, Kurtosis = 0.9151702374814694 Visual Represnetation of skeweness : For the following figure the more points are aligned with the straight line the better it is the approach for reducing skewness of PM2.5. After studying the skewness I decided to not opt for any transformation. as it shifted the PM2.5 variation interval. However I worked around handling outliers using the winsorize module.