2

I have done an experiment on fake news detection. Now I want to use other approaches' experiment results for comparison using the same data set. Do I need to redo their experiment to obtain the experiment results? Or can I use their experiment results in their paper with citations? If so, please leave the relevant rules to let me refer to.

Sorry for my confusing writing. In my research, I use machine learning to predict fake news. There are some metrics to evaluate machine learning performance, such as accuracy, precision, recall, and F1-score. What I want to do is compare our model's performance to the others(on the same dataset and the same ratio of train set and test set). Can I use their experiment results, which are accuracy, precision, recall, and F1-score, in their paper with the citations?

3
  • Can you try to clarify your question? I am having a hard time understanding what you mean. Jul 27 at 5:41
  • 2
    The rules in academia are not codified like laws. Mainly because there is no body, like a parliament, who has the authority to codify such rules. Many disciplines, but not all, have associations who have rules concerning ethics and best practice. However, many would not be specific enough for your question. Jul 27 at 7:42
  • The title question asks about using their data. The body asks about using their results. These are not the same. What do you want to do, actually? (a) redo your own experiment on their data, (b) compare your results with theirs, (c) other.
    – Buffy
    Jul 28 at 19:45

2 Answers 2

5

Of course you can compare to their data. I would even go as far as to say: you should compare your results to published data. That is the entire reason that scientific research is published.

Obviously you should cite all the papers you compare to, and make it absolutely clear where the data comes from.

3
  • Absolutely correct, but I think OP is seeking suggestions on reproducing others' experiments vs using their published results for comparison, assuming that both options are possible. Jul 27 at 13:52
  • @AppliedAcademic, I don't read it as reproducing the other's experiment but as extending their own.
    – Buffy
    Jul 28 at 19:41
  • @Buffy- OP stated, "Do I need to redo their experiment...", hence my interpretation. Jul 29 at 9:40
1

I can't answer confidently without knowing more about the situation, but as a general principle, using their published results for comparison is entirely appropriate. Redoing their experiment is also appropriate, but you just need to think about the cost/benefit tradeoff.

If their published results contain all the information you need and you're not worried about their correctness, then redoing their experiment probably doesn't provide much benefit to you. I would probably start out by using the results given, and then consider redoing their experiment if/when you run into some specific issue with it. If you do need to redo their experiment for some reason, you might want to start out by getting in touch with the authors of the original paper, because there's always more to an experiment than can be fit into the final published paper(s).

5
  • Comparing results is not the same thing as reusing data.
    – Buffy
    Jul 28 at 19:40
  • @Buffy, true, but I'm not totally sure what you mean. Jul 28 at 19:53
  • The data is used to obtain results. But the data itself isn't the "results". An experiment attempts to decipher meaning from the data. That meaning is the "results".
    – Buffy
    Jul 28 at 19:55
  • @Buffy, I don't think those words have such a cut-and-dried meanings. Typically I would say that an experiment produces "the data" and then some combination of data analysis and human interpretation produces meaning Aug 4 at 19:59
  • In numerical/computational work, the line between "experiment" and "analysis" can get a bit blurry, but doesn't sound like the (already publicly available) input data set is the data they're asking about. Aug 4 at 20:08

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .