Generate multiple dataset

[g1lai]

Obviously the collabs will save me enormous time in programming.

[TomGutman]

Your condition for the continue statement compares a value to itself. The body of the loop never gets executed.

You start jj at ii+1. Hence ii cannot be as large as the largest possible index.

Do you really want to output just the single YY and XX values, at the index values last used? Or do you want to return the entire matrices?

You have both positive and negative currents. The ratio can be negative, resulting in the log of a negative value, which is complex. Valid in a mathematical sense, but probably nonsensical in your context.

What do you propose to do with the triangular matrices, with a lot of uncalculated zero elements? There may also be a few holes in the nominally calculated part of the matrices, where you had identical currents at two different voltages.

Your pooposed regression seems invalid on its face. While you don't have it as an exact measurement point, the current can go to zero. That means that Y can go to infinity (positive or negative, depending on which current is zero). But X does not.
__________________
� � � � Tom Gutman

[g1lai]

>Do you really want to output
>just the single YY and XX
>values, at the index values
>last used? Or do you want to
>return the entire matrices?

I want to output the single YY and XX, then do linear regression of YY w.r.t XX based on Eq. 1.

>
>You have both positive and
>negative currents. The ratio
>can be negative, resulting in
>the log of a negative value,
>which is complex. Valid in a
>mathematical sense, but
>probably nonsensical in your
>context.

Yes. I need a if-statement to get rid of positive current.

>
>What do you propose to do with
>the triangular matrices, with
>a lot of uncalculated zero
>elements? There may also be a
>few holes in the nominally
>calculated part of the
>matrices, where you had
>identical currents at two
>different voltages.

Perhaps it is OK if the output is a single YY and XX ???

[TomGutman]

You can't do a regression of a single value pair. By definition, regression requires a set of data.

IAC, your data are nowhere close to meeting the assumptions that were used to generate that equation in the first place. It's a nice enough equation, in its place. It's just completely irrelevant to your data.
__________________
� � � � Tom Gutman

[ptc-1368288]

On 7/17/2009 2:59:36 PM, g1lai wrote:
>Obviously the collabs will
>save me enormous time in
>programming.
____________________________

Maybe not, not knowing what kind of statistical analysis you are looking for. My work sheet terminates with an export, and its from here that you have to describe what you are attempting. I can't make any sense of your work sheet... "linear regression", linear regression of what ?

jmG

[g1lai]

Thanks a lot for help. I understand this approach is good for dark I-V. Unfortunately, dark measurement is not possible at this moment. It good to know that I am wrong.

[ptc-1368288]

On 7/17/2009 2:59:36 PM, g1lai wrote:
>Obviously the collabs will
>save me enormous time in
>programming.
_______________________________

The calculation start with (I0,V0)=the 1st data point in the I-V curve, and derive (n-1) pairs of X-Y data; Then the 2nd data point in the I-V curve, and derive (n-2) additional pairs of X-Y data.... Finally, n(n-1)/2 pairs of X-Y data will be generated for the linear regression (Eq.1)
..........................


==> I have concluded the project last night :
http://collab.mathsoft.com/read?126038,11
...concluded the the project with an export of the I(v) function. Just collect what you want from the export, export it meshed finer.
==> don't understand "Finally, n(n-1)/2 pairs of X-Y data will be generated for the linear regression (Eq.1)". ... n(n-1)/2 mans a � decimation, then what about a � export ? Linear regression is linear interpolation, but why interpolate if you can generate the data you want in the range you want and meshed at will. It looks like you want to linearly interpolate a function. That's OK if you just have a pocket calculator and a data table, but then why about 100's visits in a Mathcad collab. Maybe you just need to learn how to interpolate on a minimum pocket calculator, not the grocery calculator from the housewife.

jmG

[ptc-1368288]

On 7/17/2009 2:59:36 PM, g1lai wrote:
>Obviously the collabs will
>save me enormous time in
>programming.
_____________________________________

I guess that one is already lost in the blue. You were using the crappy data from the original, instead of the fitted that you had from my last posting. Taking XX(1492) with the data from function, you get logY as complex, what would that mean ? and plotting 2144 traces on the Y axes... Oh ! my God, and they only plot complex

>Here's a scatter plot of the purported data. You're going to try for a linear regression of that ??? <<br>
"testI", "testV" are much less visible than X, Y.
I have made the remark before, that you persist ignoring, up yo you. This collab is so general, and universal that if you refuse helping yourself in making your stuff visible, most collabs won't even read. What do you mean by "linear regression" on that complex logy and real logX ?
Do you mean some plot in between those 2144 traces ?

jmG

[ptc-1368288]

... suite

On 7/17/2009 2:59:36 PM, g1lai wrote:
>Obviously the collabs will save
>me enormous time in programming.
________________________________

Could it be that you had confused yourself with a log plot from unknown source and that you interpreted to be log(Y/X) in the program ? If instead of log(Y/X), simply Y/X produces no more complex and the plots might make sense. But there is a curious repeat of a single negative value every 64 points. How does it arrive in the game ? Looking at the vector logY, it looks like an eye diagram. So, it might mean to pack this long vector into an array and also the logX vector. That will provide a paired data table, which paired data table could be processed directly in the program.

Pursuing further, the program generates that specific undesirable negative value, and can't be predicted where about, you can check that by varying k and see some plot to split. So, the program is not robust enough even at full numerical accuracy and full TOL ... see the split for k = 3, 4

Hope more collabs look at it.

jmG

[g1lai]

On 7/18/2009 11:59:13 PM, jmG wrote:
>... suite
>
No suite please. I have replied Tom's post that the model worked good for dark I-V while I have light I-V data in the worksheet. Will collect dark I-V data then put in your models.

My computer crashed last Friday and I am still trying to figure out how to make programs work. Suite!

[ptc-1368288]

On 7/20/2009 1:12:07 PM, g1lai wrote:
>On 7/18/2009 11:59:13 PM, jmG wrote:
>>... suite
>>
>No suite please. I have replied Tom's
>post that the model worked good for dark
>I-V while I have light I-V data in the
>worksheet. Will collect dark I-V data
>then put in your models.
>
>My computer crashed last Friday and I am
>still trying to figure out how to make
>programs work. Suite!
______________________

So, that's where it stands, complex values out of the blue [Unk353], that you can easily get rid of [Unk354], and that you just have to hint what you mean by "linear regression" and "logI-V fitting".

>No suite please< ... noticed = No suite !

jmG