Search Results  Recent posts as of less than a minute ago 
Showing 1  30 (133 results found)


Reporting a bug with drawing arrows (version 2.8.1) on Mac OS Mojave 10.14 (PC on Windows 10 doesn't have this bug). If any arrow is positioned horizontally with the head on the left and parallel to the bottom axis, the line seems to disappear (see attached screenshot). Please take action. 

I am the only one demanding... It would be really appreciated a feature regarding exporting reports from fit plot. What I am doing is, fit curves in magic plot > get equation > paste on mathematica > treat there > import results to magic plot. The slooowest for me is this part "get equation > paste on mathematica". Because: 1) I have to fit every plot and after copy only the desired content to mathematica 2) equation is not written in mathematica format (e.g. "ln" in magic plot is "Log" in mathematica, exp is Exp, () is [], etc)


Just a note, the collapse all function works properly on Figure Properties popup but not on the main menu 

Hello Lautibah,
Thank you for mentioning, yes, the new version copies the the source curve parent Fit Plot even in case of pasting on Fit Plot (so that the only formula/curve type and parameters are needed). We'll fix it later. 

Hello Again. With the new magicplot 2.8, it is not possible to copy only the fit plot curve from other file. Previously it was possible but now it copy also the table and figure from other files.
Just to let it know 

Hello,
The limit for opening files, column number and batch processing is only in memory consumption. So it depends on your current hardware and operating system, you can find the actual limit only with experiment. The program will show an error message if there will be not sufficient memory for operation.
To export multiple figures, select figures in project tree (using Shift or Ctrl as usual) and click on export in context menu. 

So, I can generate multiple figures using batch processing.
How many? If the limit is the number of columns in the text, what is the maximum number of columns allowed?
And, having imported some number of text files, can I (in batch mode) export the resulting plots to separate png files? 

When you add import multiple files to a folder, they will be sorted in alphanumeric order. example: Files (time): 10s, 20s, 30s, 60s, 120s, 1000s will be imported: 10s, 120s, 1000s, 20s, 30s, 60s
Sadly, there was a trick which worked: you could import first all the files with 2 digits, after all files with 3 digits, after ... Then you could get the files in order.
Now this mechanism won't work. Atleast the new feature let you sort it by hand, but it feels tired when there are more than 100 files to sort
Sorry if this "feature request" is too much. 

The new feature (multifolders) won't work properly with the "collapse all" function. Last child in the folder's array will collapse but the others won't. (In conclusion, all folders without a folder inside will collapse, the others won't) Greetings 

Yes, the forum lib is broken and it's not supported for a long time, so we need to move to another platform some time. 

Yes, I can send emails to all forum members.
Btw. I received no emails 

Hello!!
Thanks for the new update. I was anxious to being able to sort the folders! 

Hiho, another idea would be to implement the possibility to add new files in an existing already completed batch process. So e.g. if one fits some data during measurment to add new data without refitting everything or end up with several batch proceedings for subgroups of data. 

The answer is in mathematics: Reduced Sum of Squares (RSS aka Chi^2) which is minimized in leastsquares algorithm (both linear and nonlinear) is less for that "incorrect" curve than for theoretical one for this "noisy" data (618 vs 670). (You can find RSS in fit report for fitting with your theoretical curve while locking both parameters: the program just find these values but will not wary anything.)
Note that you are adding "noise" to X, not to Y, and that noise is not random at all.
Transforming the data is not a simple task because you need to carefully transform data points errors also (or the weight of each point). I suggest this book: http://www.facm.ucl.ac.be/intranet/books/statistics/PrismRegressionBook.unlocked.pdf 

Hello!
Thanks for your response. I am not really into stadistics so I know not much about it. I thought the calculated parameters for the fit log would be exactly the same as the exponential and, finding out other response made me concern about if my previous fitting graphs with Magic Plot were OK or not. You were right, in excel I added 51 rows instead of 52, but the log and exponential curves have the same R^2 and parameters (but as you explained thats because of a non linear fitting).
I am convinced about what you said, but it blows my mind how it's possible how the RSS is less by fitting "worse". Probably as you said, it is because it's a linear fitting.
For example, I have entered this new formula "if(Col(A)%2==0,Col(B)+0.5, Col(B)  0.5)" (with the exception of row 1). Then, the "sum" of each point is cancelled with the other in each row, so the fit curve should be the same as fiting with Col(B) directly. I am attaching the figures below and the mppz (it only contains the new points). In the figures you'll be able to see that the fit curve matches with the theorical (i didn't draw it, but its exactly drawn in the middle of the points). In the case of the exponential curve, it diverges from the theorical curve, and the parameters are just crazy: theorically a = 3, b= 4 Fit in the exponential curve a = 1.25+0.79 (63 %!), b = 4.5+0.2 (not bad)
I am sorry for this post, I am not blamming you or whatever, I am trying to understand if it is more interesting to convert my data into Log curves and fit them than fiting with an exponential (looks like linear fitting is worse in this case)
Greetings!
Edit: axe labels are not 100% correct, and fitsum is just the fit curve 

Hello Krairy
No, MagicPlot does not support multidimentional fitting yet. 

Hello Lautibah,
There is no error in MagicPlot fitting algorithm. First, R^2 0.9995 and 0.9999 can be treat the same value for nonlinear fitting algorithm, the tolerance is enough. Second, the comparison is not so correct. You have 51 (not 52) rows in Excel file. Also Excel uses not universal LINEAR fitting (the model function linearly depends on varying parameters a, b) and this algorithm is not iterative, it's just a formula, so it can provide more tolerance.
Then, 10% standard deviation of parameters found is not linked to residual sum of squares (RSS) and R squared. The standard deviation of parameter shows in fact how much the RSS will change (relatively) if parameter is varied. Imagine that you fit exponential data with a line. You may get huge RSS because line does not fit well, but you may have less than 10% std. dev. because "there is a single way to draw a line". 

I've done the same on Excel and it is fitted with the same R^2, expected... wow
Something wrong with magic plot? 
bug fit? 
Posted by: Lautibah Date Posted: July 23, 2018, 9:39pm 
WordHits 1 (100.00%) 

Hello.
Two things: 1) I've found that if a column has a formula that depends of "i", when you add a new row it won't be autorecalculated even if the column has that feature activated.
2) I am sorry, I know not much about fittings, but I've seen something quite strange. Below I am attaching an .mppz file in which I've done the following steps:  create new column with formula "i" renamed as "value"  create a new column with formula 3+4*log(i)  create a new column copying all the data but one, that is modified.  Create fit plots of each two columns vs value  Add custom curve "a+b*log(x)". In the case of unmodified data, R^2 = 1 as expected. In the case of modified data R^2 = 0.9999 (ok, makes sense)
After:  Create fit plots of each value vs columns. (With this, the graph will be exponential instead of logarithmic)  Add custom curve "exp((xa)/b)". In the case of unmodified data, R^2 = 1 as expected. BUT, in the case of modified data R^2 = 0.9995 . Moreover, the difference of the parameters in comparison of the true value is quite greater
Why?
Isn't the data the same? Why is it the R^2 different?
I've found this because on my true data, there is a difference in R^2 of about 0.1 and that surprised me (parameters differs in 10%...).
Greetings 

Hi All,
Is there any possibility to deconvolute multidimensional signal in MagicPlot?
Thanks! 

I think it could be interesting to define the range for a variable in the fit plot. (I know some option is to set a fixed value, fit, unlock the variable and refit. Not always is that cool)
Thanks 

Thanks, worked like a charm. 

Hi! To begin with MagicPlot cannot handle such double construction: 3.9<x<6.8, and it is returned always true. Secondly, you have only 2 points so you need 2 if's, like this:
Code
if(x<3.9, k1*x+c1, if(x<6.8, k2*x+c2, k3*x+c3)) 



Hi Alex, can you please help. I need to fit three part straight lines to a data. I can fit first two with the given formula but the third one seems not possible.
if(x<3.9,k1*x+c1,if(3.9<x<6.8,k2*x+c2,if(x>6.8,k3*x+c3,0)))


Hello,
No, only numbers are supported on axis labels currently and it is not planned in upcoming release. 

Hello. Sorry for my high activity in the forum.
Is there a way to set axis labels as text? For example: instead of numeric values, text values of a column in a table.
Thanks 

Thanks! I also had that idea, but it turns out a problem when importing multiple files with different columns.
Since a new update I'm using the 1.0E999! Thank you 

Hello,
It seems to be a bug, the "NaN" text must be recognized as numeric column, we'll fix that. Thank you finding!
In current version you can only set the column type manually in context menu on preview table header, see the screenshot attached. 
Showing 1  30 (133 results found)

