https://bugs.freedesktop.org/show_bug.cgi?id=87790

--- Comment #6 from mencaraglia <[email protected]> ---
What I expect is 0,7364; In these days ha have done some other checks; using
the spreadsheet GNUMERIC for instance I find for r2 the value 0,98.... (same as
libreoffice); same value using the Google sheet. Using the spreadsheet SCIDAVIS
the R2 is 0,73...,
Using the program GRETL the results are 0.98... if I use the ordinary least
square methd (fit function a *x ) and I find 0,73.... if I use the non linear
fit (fit funztion a * x).

Looking for other occurrences of the R2 problem (google may help) I found that
several years ago ( 2006 ) the same problem (wrong values of r2) had been found
for excel and the origin was in the linest() function.   see for instance
http://support.microsoft.com/kb/829249

I have not worked any more on the subject since (a) i had a work around and (b)
it's the first time I fit with Y = a *x  and probably next time will be in 10
years (;-) ) 


Here below please find the two results obtained from GRETL and the one from
SCIDAVIS

Modello 1: OLS, usando le osservazioni 1-50
Variabile dipendente: v2

             coefficiente   errore std.   rapporto t   p-value 
  -------------------------------------------------------------
  v1           −1,21416      0,0195447      −62,12     2,91e-48 ***

Media var. dipendente  −3,693268   SQM var. dipendente     0,833732
Somma quadr. residui    8,977973   E.S. della regressione  0,428047
R-quadro                0,987462   R-quadro corretto       0,987462
F(1, 49)                3859,178   P-value(F)              2,91e-48
Log-verosimiglianza    −28,01571   Criterio di Akaike      58,03141
Criterio di Schwarz     59,94343   Hannan-Quinn            58,75952
Note: SQM = scarto quadratico medio; E.S. = errore standard


----------------------------------


Sono state usate derivate analitiche
Tolleranza = 1,81899e-12

Convergenza raggiunta dopo 4 iterazioni

Modello 2: NLS, usando le osservazioni 1-50
v2 = alpha * v1

              stima     errore std.   rapporto t   p-value 
  ---------------------------------------------------------
  alpha      −1,21416    0,0195447      −62,12     2,91e-48 ***

Media var. dipendente  −3,693268   SQM var. dipendente     0,833732
Somma quadr. residui    8,977973   E.S. della regressione  0,428047
R-quadro                0,736410   R-quadro corretto       0,736410
Log-verosimiglianza    −28,01571   Criterio di Akaike      58,03141
Criterio di Schwarz     59,94343   Hannan-Quinn            58,75952
Note: SQM = scarto quadratico medio; E.S. = errore standard


----------------------------------------------------

[02/01/15 14:31    Plot: ''Graph1'']
Non-linear fit of dataset: Table1_2, using function: a*x
Y standard errors: Unknown
Scaled Levenberg-Marquardt algorithm with tolerance = 0,0001
>From x = 0 to x = 3,912
a  = -1,21415873230445 +/- 0,0195446571501494
--------------------------------------------------------------------------------------
Chi^2/doF = 0,183223940915315
R^2 = 0,736410071654573
---------------------------------------------------------------------------------------
Iterations = 1
Status = success
---------------------------------------------------------------------------------------

-- 
You are receiving this mail because:
You are the assignee for the bug.
_______________________________________________
Libreoffice-bugs mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/libreoffice-bugs

Reply via email to