I'm sure you got tons on your plate Eugene, but if time avails and it deems practical to for some of the following to do, it would be great to have the ability to further customize "analysis series" such as :
. Change scale of scatter plot.
. Have the ability to generate linear or quadratic equations with respective RSquared, ANOVA, F and T statistics.
. have the ability to sort independent variable so as to include/ exclude portions of the range, or perhaps just the ability to remove outliers.


This is just some ideas to consider. Realizing your time is limited, or this is outside your scope, can users build open source GUIs for the above purpose or for other purposes that help us analyze data? This was kind of what I had in mind for an R interface, but if data analysis and backtesting can be housed in the same program, I imagine it would make analysis far more efficient.

OK thanks for everything. This will be my last post for now. Cheers!! :)

Mike, having reviewed your ideas, let me kindly note that my vote would be for this good thing of Glitch to stay simple and clear as it has always been. Because with what you suggest, it will be something completely different from the initial design.

1 - The scale of scatter plot - what for? This visualizer doesn't have any idea of time. Simply re-run your strategy on a different bar scale.
2 - I haven't heard many of these words. ;) I know, I know - how can a human live without knowing what's ANOVA, ARIMA or GARCH but that's a different story.
3 - What's the purpose of being able to manipulate the range? What is the point of removing outliers - for some reason it doesn't seem obvious to me?

QUOTE:
can users build open source GUIs for the above purpose or for other purposes that help us analyze data?

Sure - the Performance Visualizer API is open, an example source code was posted to the Wiki. Can't see why one can't connect to R via COM interop in a visualizer to analyze data.

What I mean by scale relates specifically to the range of values a given data series displays on the X axis. Now for example we look at a 3 day ROC plotted against the Y, Percent profit, we will see the densest cluster of values reside near it's mean; however there will be points all the way to the extremes. So say the cluster resides near +3% ( in a breakout strategy)or -3% in a dip buyers strategy, we may miss good visual descriptive information by the fact that our scatter plot incorporates values of say -99% or 500%. We may want to scale down into where most the distribution is to see how things look closer in.

Also there have been too many times I have become ecstatic doing a scatter diagram or a regression in excel, generating what seems very meaningful R squared values, only latter to realize that this value was true based on the last 3 points in the series ( or whatever small % of data ). Removing these outliers reveals the true relationship for the majority of the data, which is statistically random/ independent.

Having the ability to zoom in, if you will, allows us to really see the data. Removing outliers and or parts of the ranges gives us hope of describing the specific nature of the distributions or at least not have the data "tell" us what seems meaningful, only to realize it was based on a few points. Through my investigations, it is quite clear that data is most often not linear or normally distributed( heteroscadastic sp?. In order to understand what our data tells us requires having the ability to zoom in or scale out as required.

Thank you for the in-depth explanation that made it clear for me.

1. An option to remove outliers would be quite involved to implement, and complicating this code further is an undesirable thing. Sorry but I would like to keep from implementing it or any other involved change in a Fidelity supported extension.

2. I've tried implementing zooming in/out in Analysis Series but probably due to some internal differences from other (community) visualizers of the same kind, this feature isn't working as expected. If at some point in future I'll have the necessary inspiration, the issue may be fixed.

Very well then. I understand that implementation of such methods remain risky from a programing and thus a support perspective. Perhaps then an efficient integration of WL with R might be desirable for my practices.

Anyway, sorry to push the envelope, but having not much time studying programing, it is difficult to understand what is practical from a programmer's perspective. Figure it doesn't hurt to ask though.

Have a nice weekend Eugene.

Hi Mike,

just a quick reminder that you can do all these things in excel.

Cheers,
Bruce

Hi there Bruce. Hope you are well.

I have found Excel to have it's limits with large data sets; unless however there are versions that I don't know of yet. Excel seems to grind after 20,000 or so. It does however prove a good adjunct to my research.

Thanks.

Cone, the Analysis Series looks excellent. I am amazed I haven't used it after about 15 years.

Is there a way to fine tune the X-axis? For example, on the attached chart, my trade count drops off massively when ROC50 goes from 0 to 25% (ie the first to second points on the red line). Is there a way to effectively zoom in on this range (eg 0-25 in steps of 5) to fine tune a filter?

Thanks and regards
Rod

Thanks Rod.

No, there isn't a way to "zoom in". The histogram bins are fixed. In addition, even if it were possible the Y axis step would be too small.