Emil O W Kirkegaard

I would like to see more studies done, but bizarrely this is not something many people have studied & published! Science is broken and we

It's not ideal for anything except very simple stuff.

>People using Excel for papers >In the current year.

You would be surprised... And then we have evidence showing that errors in spreadsheets are disturbingly common. http://arxiv.org/abs/0802.3457 

Lulz. I thought people would have learned after the Rogoff-Reinhart affaire. It is surprising because using a proper script makes things

Clearer, reproducible, and reduces probability of error. The only reason not to do it is because one doesn't want to learn how to write code

And that's a bad attitude from an academic. Surely if one has the brains to understand the concept of significance testing, one can code.

Sloppy methods won't disappear by coding, but the issue with Excel is that it makes it easier to have an error hiding in a cell

That you forgot about,and coding it makes it easier to see what your analysis flow is actually doing

I like to do defensive programming with assertations for expected results (e.g. no missing data after imputation) https://en.wikipedia.org/wiki/Assertion_(software_development) …

Heard about the infamous Reinhart & Rogoff paper where big policy conclusions where drawn with errors in their Excel sheets?

Of course. But I rather cite statistics than case studies.

Sure, but for persuasion purposes a good case study might work wonders.

Don't think anyone was saying Excel is bad though :) (And their most grievous error was bad statistics, the Excel one is what ppl remember)