I am dealing with the various climate proxy data (e.g. geochemistry, pollen, isotopes, etc.) from various locations. I would like to compare statistically (component analysis) all the proxy data, but the problem is missing points.
I am not an expert in the statistical analysis. Is it possible to fill the missing values with the median values and then do a component analysis?
Any help would be greatly appreciated.
Thanks
Hi Praveen,
there are many options to tackle missing values. Regarding climate data, I observe that case deletion and median are widely used. Data imputation using more sophisticated methods are also adopted, such as autoencoders ANN or evolutionary algorithms.
Perhaps, one of the attached papers can help you a little bit.
Regards,
Fábio
Conference Paper An Evolutionary Missing Data Imputation Method for Pattern C...
Article Multi-Objective Genetic Algorithm For Missing Data Imputation
Conference Paper Time series imputation using genetic programming and Lagrang...
as suggested above, impute the means or leave out the records with missing values.
if you use more sophisticated techniques such as rpart, surrogates will be used automatically (to form tree leaves and partitions) for missing values.
- Badges
- Science topic