If it not follow the bellcurve, should we change the lenght of the Oscillator?
Dicargo_Beam
⋅
@matquykyo811, No matter how much you adjust the period value, there will be some oscillators that do not follow the normal distribution according to the calculation formula. I know that the indicator values for machine learning should follow a normal distribution. I built this classifier to discriminate against such indicators.
matquykyo811
⋅
@Dicargo_Beam, if so, is there any functino to convert an array to a new array that follow the normal distribution, I hear something about john ehlers function but still looking for!!!
capableProfile60219
⋅
Hi. First of all congratulations for this indicator which I find very useful for any type of research but, is not clear to me such a large std, with values eg. 22 or 20: What does such a large standard deviation refer to?
Dicargo_Beam
⋅
@capableProfile60219, A high std results in a wide confidence interval. When the average is 50 and the std is 10, the typical oversold and overbought points are 30 and 70, which has a 95% confidence interval (this means that 95% of the values in the sample fall between 30 and 70). is 20, the 95% confidence interval is 10, 90 (this means that 95% of the values in the sample are distributed between 10 and 90). Considering 'RSI 14' as the most basic and ideal (std 10), std of 20 means that it is not good.
capableProfile60219
⋅
@Dicargo_Beam, lot thanks, clearly your way. Best would be an input in your indicator where there is choice of bar numbers for the test, what is your opinion ?
@Dicargo_Beam, Ok. I try, my english not good anyway. In text box of indicator at work appear, seems to me, the numbers of bars processed at the end of gaussian research. Do you think that insert the number of that bars in input should be in process a further aid ? Thanks again.
Dicargo_Beam
⋅
@capableProfile60219, I thought that the more samples there are, the higher the accuracy, so I added all of the bars starting from the leftmost bar on the chart as samples.
capableProfile60219
⋅
@Dicargo_Beam, Thank for your reply. Of course more population more precision, less error. I'll should be only curious to see what be happen for example in a session compared to another o a previuos and in different timeframe with different numbers of bars. My interest is that this brillant indicator could be a valid substitute for several complex and questionable indicators on the identification of the dominant cycle bar by bar.