| T O P I C R E V I E W |
| erickzhou |
Posted - 07/14/2010 : 05:04:06 AM
Origin Ver.8 and Service Release (Select Help-->About Origin):
Operating System:windows XP
I used "nolinear curve fitting" to fit my data(256 points) with the equation of "y=A+Bx+Cx^2". I want to get the parameters A, B and C.
The results are shown as follows
value standard error
A 2305.62049 0.67257
B -0.36728 0.0243
C -2.87 0.0243
I want to know what is the meaning of "standard error", is it the smaller the better?
If I have another results as follows:
value standard error
A 2312.62049 0.13257
B -0.34728 0.0143
C -2.97 0.0143
Can I compare the above two results by the absolute value of "standard error", such as 0.67257 and 0.13257,or perhaps I should compare the results with relative value of "standard error", such as the quotient as 0.67257/2305.62049 and 0.13257/2312.62049.
I will be very appreciate for your reply, thank you!
|
| 2 L A T E S T R E P L I E S (Newest First) |
| erickzhou |
Posted - 07/15/2010 : 08:43:34 AM
Thank you for your reply! |
| larry_lan |
Posted - 07/14/2010 : 11:13:05 PM
Hi:
The standard error of a parameter ca be used to computed t value, p value, etc. These values suggest if the parameter is significant different from 0.
Standard error of parameters are not quantitative value, so, you cannot compare the parameter standard error between two fit. It's qualitative value and you can compare the standard error with the fitted value.
You can be told from the parameter standard errors that how precision of the fitted values. These value give you a primary impression on goodness of the fitted parameters. Usually, we would like to see the magnitude of the standard error values are less than the fitted values (0.13257 vs. 2312.62049 in your case). If these values much larger than the fitted values, your fitting model may be overparameterized.
However, you can compare two fit by their R value, R square value, RSS, etc. This page maybe helpful, and you can also look for some statistics text books for more information.
Thanks
Larry
OriginLab Technical Services |
|
|
