Another approach to using agreement indices is to use outlier analysis. If the outlier analysis is used, the agreement index is calculated (see details below). This is then considered an acceptance threshold for the various contractual indices. For this agreement index, a calculated value is calculated if you dispute a value for a combination (combination) or a match (D_Sequence). This value is about 100% if the item surveyed is combined and expected and decreases relative to the probability, if the combination is not very likely. The value of these can also increase by more than 100% if the agreement is exceptionally good. At this threshold, we can then calculate the logarithmic average of each agreement index that composes it. This is given by: This overall chord function has some interesting features. The first can be found taking into account the particular case of combinations of probability distributions (performed here with the combination and D_Sequence): in such cases, errors are not independent, because all comparisons are made with the same rear distribution that has an error that decreases with the square root of n.
The particular case of Gaussian distribution combinations (generated by C_Date) yields identical results to Gauss`s direct combinations (with C_Combine) and it therefore seems reasonable that the threshold for acceptance of the combination should be the same as the chi squared test normally performed. It turns out (and this can be verified by trying groups of values) that the threshold for Aoverall, which corresponds to the chi-square at 5%, is the same: the indexes of agreement tell us something about how the previous model matches the observations (expressed in terms of probability). This is important because it is very easy to build a model that is clearly at odds with the observational data. There are four forms of the agreement index calculated by the program: For simple combinations generated from the Combine function (), D_Sequence or operator-operator, there is only one independent parameter. In these cases, Foverall is equal to fdel. As in this case there is only one independent parameter in the comparison model, we give a particular name to the agreement index in this case: the value of A`n is about 60% for a wide range of values on n. This is the justification to take 60% as a threshold for the acceptance of certain contractual indices. The argument in favour of the same threshold for Aoverall and Amodel is that we expect ln (F) to happen to go 0 if we increase the number of parameters. This threshold is the symbol: 100% differences have the same meaning as for different chords. The mathematical wording here is not entirely strict, and given the nature of the problem, it is probably inevitable.