Hello Christopher Brooke ,

If you have missing values in your data set, you have 2 choices before you run the clustering process. First choice is that you can delete rows, which have missing value, from dataset. And the second one is you can use any missing value estimation method. You can find different estimation methods about missing values. Such as: replace with mean, replace with mode, replice with median, etc. So if you don't want remove missing rows, you should specify an estimation method and apply your dataset. And then run k-means or any other clustering algorithm.

Best regards.

You can use mean value for the missing data. That's what I did for my cluster analysis using SPSS.

Thanks for the input, knnImputation with the cluster means seems to have worked well.

The rase way is to added through the use of a probabilistic model

You can find here some additional and useful suggestions: https://www.displayr.com/5-ways-deal-missing-data-cluster-analysis/

Recommend using osample, or esample - try running entropy matching with ebalance in stata

I think for any case with missing data, the key, at least for me, is to understand why the data is missing in the first place. Is there a pattern? Is the "missingness" seemingly random, or is there a process which we can define to some degree that creates the missing observations. As noted in the other answers, there are many approaches to missing data, but choosing one over the other rests (typically) on different set of assumptions, and that again brings me to the fundamental question of why the data is missing in the first place.

Sorry, I thought I finished but realized hit add before finishing. Of course not my area, but I am guessing there are specific mechanisms wherein some species are more likely to be missing than the others. If those causes are known, they can be used, in a modeling-based imputation method. In my experience, imputation has a crafty side in it, and blindly choosing one method over another based on some purely exogenous mathematical criteria generally does not do the trick.