I manually downloaded as cvs files from CRRO before asking Claude to do the following:
Prompt 1:
Extract the weight data from these four csv files. Calculate the median and mean for each coin issue, as well as the relative standard deviation. Create a box and whiskers graph to allow visual comparison of the four issues.
Prompt 2:
Same task. Small tweaks. Remove any outliers over 12g and under 4g. Remove the files names just use the issue names, do not included dates or mints.
This is a task I could very easily do without Claude. The advantage was only time. I wanted to fact check a sentence in a co-authored piece. I didn’t want to take the time to do the manual data cleaning and manipulation. Claude continues to try to be ‘helpful’ but this unwanted ‘help’ is very good for me to identify my own superior expertise and knowledge. I can see how this could gaslight someone into thinking they know more than they really do. It is also a reflection of how much outdated information is encoded in our linked open data, from revised dating to mint locations, etc etc.
I also like the scatter added to the traditional box and whiskers. The messiness is an advantage for seeing the real nature of the data.
Ok gotta do the school run.
