I'm going to improve my result. I know how to calculate the mae and mare by formula. But, some papers, they showed the additional result of MAE and MARE in bracket as shown in the picture. For instance, the value of (0.081) and (0.186) for MAE and MARE respectively. Could anyone advice me :
Thanks in advance
I think because it says 500 replications, that the value in brackets is the standard deviation of all 500 MAE results.
So to calculate it, you’ll have to repeat your simulation 500 times and calculate 500x MAE values, then calculate the standard deviation from those.
Here is the Wiki on standard deviation, if you’re unfamiliar with it.
https://en.m.wikipedia.org/wiki/Standard_deviation
A larger standard deviation means that there is a wider dispersion of MAE reuslts for the tests. A smaller standard deviation would mean your tests are consistent in producing the error. Smaller would be better.
i hope that helps!
Dear Dr Jamie Matthew,
Thanks a million for your comment.
Does it important to provide the standard deviation result since the error evaluation already shown which method has the best performance.
For instance, for 10% missing data percentage in MAE, the DA outperformed the rest methods since the smallest is the best.
You’re welcome!
i suppose it is important if reporting it is the standard in your research field?
The only things that I infer from the std devs reported in the table is that the MI method produces less consistent values compared to the other two, and that the EM and DA have very similar consistency with repetitions.
I dont know what any any of these methods are, so I couldn’t say whether the std dev is a useful metric or not, other than an extra comparative test between the 3 methods.
Dear Dr Jamie,
It is indeed true. The result of SE of MI method are inconsistent and arguable.
Thank you very much
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Software
- R Statistical Package