My Math Forum (http://mymathforum.com/math-forums.php)
-   Algebra (http://mymathforum.com/algebra/)
-   -   How to calculate max/min scales on a scatter plot (http://mymathforum.com/algebra/42328-how-calculate-max-min-scales-scatter-plot.html)

 expertalmost March 24th, 2014 09:20 PM

How to calculate max/min scales on a scatter plot

Good morning!

I have 3 log scatter plots that I want to establish smooth maximum and minimum lines. What is the usual mathematical method to do that? (Image and excel file links below.)

The black lines on the scatter plot images are hand drawn. The third scatter plot is especially tricky and not amenable to a moving average plus stddev because of the data clumping. Note: This is time series data so new data constantly comes in. In other words, I cannot just use the whole data population in one shot.

Any ideas would be greatly appreciated.

Excel File: https://dl.dropboxusercontent.com/u/440 ... atters.xls
Image at: https://dl.dropboxusercontent.com/u/440 ... atters.jpg

 skipjack March 25th, 2014 06:13 AM

Quote:
 Originally Posted by expertalmost I want to establish smooth maximum and minimum lines. . . The black lines on the scatter plot images are hand drawn.
Is there any reason for thinking that a "black Line" that was simply a "horizontal" straight line would be any less useful than the slightly wavy line you seem to want?

 expertalmost March 25th, 2014 06:27 AM

Re: How to calculate max/min scales on a scatter plot

Thank you Skipjack for your time and efforts!

Actually, the lines I now generate are too "wavy" with too much lag. Not to mention by a rather ad hoc solution. I take the 25 largest/smallest of the last 100 elements and average that. I also add a stddev amount to the max. Then smooth by damping (multiplying the change by .1 and using it). I was hoping there was a more elegant/roboust/general solution as I have to tweak the stddev and damping factors for different data sets. Basically, I need a 0-1 envelope covering 90% of the data.

Straight lines are fine. But as it is time-series data, I can't imagine it being a straight line. But, I'm not formally trained in this so maybe it could be.

Any solution ideas?

 skipjack March 25th, 2014 07:13 AM

You imply things are changing over time (other than random variation). What is the nature of that change? If you don't know, you also don't know how to assess whether 90% of your data lies between certain values. You referred to a "0 - 1 envelope" - does that imply a mean value (in the same scale) of 1/2? If so, how can you have a mean value for data that's changing over time?

 expertalmost March 25th, 2014 07:22 AM

Re: How to calculate max/min scales on a scatter plot

Maybe I'm missing something? But don't most econometric time-series data have non-stationary means? Isn't that what I'm seeing when I take a mean on so a sample of the data and it varies?

 skipjack March 25th, 2014 08:36 AM

If you measure sea-level only at times close to high tide, the mean level will be higher than if you measure sea-level only at times close to low tide. You need a basic understanding of the tides in order to make useful statements about 90% of change in sea-level due to the weather being within such-and-such limits.

 expertalmost March 25th, 2014 07:03 PM

Re: How to calculate max/min scales on a scatter plot

Okay Skipjack. You win. I'm an idiot without a basic understanding.
Nevertheless, it does not change the fact that I am looking for a mathematical method to calculate maximum and minimum levels for scaling.
Any help is greatly appreciated.
Stupidly,
Alexander

 skipjack March 26th, 2014 08:03 AM

How does 90% come into that? Why is 90% any better than 100%? You seem to be saying you want to portray 90% of your data because you've arbitrarily decided that the remaining 10% of it doesn't matter. Before, you said you were doing some rather complicated things with your data and wanted to find something more elegant and more general. However, what constitutes 90% of your data depends on what data you select, i.e., what time interval.

 expertalmost March 26th, 2014 02:49 PM

Re: How to calculate max/min scales on a scatter plot

Forget it. It's obvious you don't want to help. You just want to argue and prove how smart you are at the expense of others. This is the last time I use this forum. And you can stop puffing yourself up skipjack. Everyone can see it's just hot air.

 All times are GMT -8. The time now is 08:59 AM.