0

I have 2 discrete data sets (responses given my 2 different sets of people) of size 6000. Let's say that each data point is either 1 or 0.

S.No. Dataset 1 Dataset 2
1 1 0
2 1 1
3 0 0
.... ..... .....
.... ..... .....
6000 0 1

I have calculated the error rate (by comparing these answers against true answers). The error rate in dataset 1 is 84% and the error rate in dataset 2 is 89%. I want to understand if the difference between these 2 error rates is significant or not. How do I do it? Please help me with this.

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
user2409011
  • 241
  • 1
  • 3
  • 6
  • 1
    What is an "error" here? – Henry Aug 14 '21 at 10:24
  • Error rate is simply the mis match rate between observed answers in these 2 populations and the actual answers. – user2409011 Aug 14 '21 at 11:07
  • Error rates of $80\%+$ are not good in a binary prediction. Does $1$ mean correct or wrong or something else? And $0$ wrong or correct or something else? – Henry Aug 14 '21 at 11:13
  • It was just an example. It could be as low as 2 or 3%. My question is more around on how to decide if there is a significant difference between 2 sets of observations. – user2409011 Aug 14 '21 at 12:58
  • Is your question different from this? https://stats.stackexchange.com/questions/113602/test-if-two-binomial-distributions-are-statistically-different-from-each-other – Dave Harris Aug 15 '21 at 16:58

0 Answers0