Time for a quick poll. Would you consider yourself capable of reproducing the results presented in the last paper you read (where you weren't one of the authors)?
Poll
Would you be able to reproduce the results presented in the last paper you read?
I was able to reproduce *some* results. But there was obviously a presentation flaw in the paper which makes me think that this paper was not that fair.
However, it was the only paper I had problem with (I implemented nearly a dozen). Other papers I implemented worked exactly the way they were supposed to. This is to speak modulo experimental parameters! Some people usually want to present their results using parameters that fit well their algorithm -- and it that case the results are OK. However, real-life parameters, when different from those of the paper, sometimes lead to very different results although the whole thing is reproducible.
My main point therefore deals with experimental parameters and fitness to the problem that is addressed in the paper. I guess every research field has to describe its own reasonable parameters...
The last paper I read is "Reproducible Research in Signal Processing". It is reproducible I think. However, most papers I have read are not reproducible. Unfortunately ,this problem is particularly serious in my country the P.R.China.
No, it was my own first paper, which I wrote before using R/sweave. I have some problems reverse-engineering it into the sweave system and still struggle to even replicate not reproduce. Another argument for RR systems...