Ranging the values

we have a time varying graph, after sampling we have a series of non-uniform values with a minimum and maximum value far away from each other. This is expected, maximum and minimum definitely can not be equal. Also it totallly defends on the scale factor that defines far awqay means how far away ?

Now we need to find a way to range these values so that the resulting graph remains with minimum and maximum of 0 to 1. what are the methods available..we have tried by dividing the entire series with the maximum value : If you want the resulting range between 0 and 1 then this is the only way.

any other methods : No if you want between 0 and 1

The problem with this method, we have the truncated the values after division, Got to do it ,and even if you don;t any way actual undivided values were themselves truncated.

so we cannot the exact orginal graph if we try to extract it from the resultant graph - why not if you truncate after a sufficient number of decimal places.

Sorry for the emotions of Ducon but the same emotions will be repated by anybody who reads the post and may have additional etc too.

Thanks sb. When I read it my emotions was but tried to control it.

If the min/max values are outliers, why not get rid of them? You could do a 2 standard deviation test to see if they fall within the 95% range of your sample couldn't you? If they don't fall in that part of your sample, bon voyage to those numbers. Does this seem valid to others?

Take the log of the time series first (to compress the dynamic range), then normalize the result to 1.0. Take a look at this URL for an example (which happens to also remove the "trend", which you may not need to do):

http://oracledmt.blogspot.com/2006/03/time-series-forecasting-2-single-step.html

Hi ...

Maybe I am over-simplifying, or maybe I didn't understand the question (used that excuse a lot ), but is seems if you normalize the data so the lowest value is 0, then so the highest value is 1, it answers the question.

Here are just some random numbers to show what I mean ...

Sorry if I missed the point .

Kind regards ...

I don't know what you mean by normalizing the data .

But data normalization is a trichy process and interpretation of the result trickier.

It is possible to do it but rarely I needed it (except may be in a few cases yes). I still say let us see the dara series in preferably Excell table pasted (can not access links to the external file-upload and get a IWSS2 message ) and let us see how we can crunch it.

But looks like the OP has lost interest (as usual with OPs they lose interest too quickly) and not posted back.

No use of guessing around with unknown data. How the hell one knows, the data may fit exactly beta or exponential distributions or so many other common (but ab-normal ) distributions ?

The OP is as usual silent, Can you please give some data it will not compromise your privacy policy since we don't have any idea what the data is about, so we can have a look on the distribution ?

An open ended questions can be at very best wrong answered.