falkao
Stalker
- Joined
- Oct 25, 2005
- Posts
- 1,993
- Location
- Italy
- Society
- Freelancer
- Avatar Name
- Marc falkao Falk
A community mining experiment
Maybe you've followed one of the mining threads with respect to our publication and related discussions.
One issue that is still open is "How random EU is?". Therefore this little experiment.
The only thing one has to do is to
drop one bomb or one probe in an area where you know there are resources (CND excluded)
and answer in the poll if you've
got a find or not; with yes (find) or no (no find).
Hence only the first probe that you drop after reading this is relevant.
Results
A total of 133 avatars did participate and 44 of them had a find. The find rate is therefore estimated as 33.1% with a 95% CI ranging from 25.7% to 41.5%. This means that the real find rate is somewhere between those limits with a probability of 95%.
The aim of this experiment was to show that a mining find is purely random triggered according to some preset find rate. In this poll we did collect finds of different avatars. If we compare the poll result with the find rate from a mining publication were only data from 2 avatars was used we do get the following:
Fig. 1: Find rate and respective 95% confidence intervals of poll and mining publication
The difference between the poll find rate of 33.1% and the 27.2% of the publication is not statistically significantly different (p = .127, Chi-Square test). This can be also seen from fig. 1 as both CIs do overlap.
Discussion
This poll has several limititations. First of all it is not possible to verify answers and hence readers not aware of the problem might have answered only for fun. Furthermore, there is also the possibility for a positive reporting bias as those getting a hit might have returned more frequently to answer the poll as those that did not. In the last 10 answers for example there have been 5-7 finds.
Nevertheless, this simple experiment did show how a poll on EF can be used to assess unknown frequencies (rates).
Since we were not able to falsify the null hypothesis (the difference between both rates was not significant), we can't automatically conclude that both rates are identically. It is however, quite improbable that the still possible difference is large. For instance, to show that the observed 33.1% from this poll is significantly different from 27.2% at least 946 answers would have been needed.
Maybe you've followed one of the mining threads with respect to our publication and related discussions.
One issue that is still open is "How random EU is?". Therefore this little experiment.
The only thing one has to do is to
drop one bomb or one probe in an area where you know there are resources (CND excluded)
and answer in the poll if you've
got a find or not; with yes (find) or no (no find).
Hence only the first probe that you drop after reading this is relevant.
Results
A total of 133 avatars did participate and 44 of them had a find. The find rate is therefore estimated as 33.1% with a 95% CI ranging from 25.7% to 41.5%. This means that the real find rate is somewhere between those limits with a probability of 95%.
The aim of this experiment was to show that a mining find is purely random triggered according to some preset find rate. In this poll we did collect finds of different avatars. If we compare the poll result with the find rate from a mining publication were only data from 2 avatars was used we do get the following:
Fig. 1: Find rate and respective 95% confidence intervals of poll and mining publication
The difference between the poll find rate of 33.1% and the 27.2% of the publication is not statistically significantly different (p = .127, Chi-Square test). This can be also seen from fig. 1 as both CIs do overlap.
Discussion
This poll has several limititations. First of all it is not possible to verify answers and hence readers not aware of the problem might have answered only for fun. Furthermore, there is also the possibility for a positive reporting bias as those getting a hit might have returned more frequently to answer the poll as those that did not. In the last 10 answers for example there have been 5-7 finds.
Nevertheless, this simple experiment did show how a poll on EF can be used to assess unknown frequencies (rates).
Since we were not able to falsify the null hypothesis (the difference between both rates was not significant), we can't automatically conclude that both rates are identically. It is however, quite improbable that the still possible difference is large. For instance, to show that the observed 33.1% from this poll is significantly different from 27.2% at least 946 answers would have been needed.
Last edited: