What are the possible sources? (preliminary discussion)
1. Errors in the model
2. Errors due to mistakes (arithmetic/programing)
3. Errors in data (noise included, lapses in recording)
4. Machine errors in calculating devices (rounding/chopping,underow/overow)
5. Errors due to truncation of formulas (eg. Taylor seriesapproximations)
Estimating errors
The exact value that we want to handle is referred as true value
(denoted by xT ). A particular value that we use to represent it is
referred as approximated value or simply as approximation (denoted
by xA).
There are several simple ways to estimate error associated
with xA.
i. Error of xA: err (xa) = Xt-Xa
ii. Absolute error of xa: Abs(Xa) = |Xt - Xa|
iii. Relative error of xa: Rel(Xa) = |Xt- Xa|/|Xt|
iv. Number of significant cant digits
1. Errors in the model
2. Errors due to mistakes (arithmetic/programing)
3. Errors in data (noise included, lapses in recording)
4. Machine errors in calculating devices (rounding/chopping,underow/overow)
5. Errors due to truncation of formulas (eg. Taylor seriesapproximations)
Estimating errors
The exact value that we want to handle is referred as true value
(denoted by xT ). A particular value that we use to represent it is
referred as approximated value or simply as approximation (denoted
by xA).
There are several simple ways to estimate error associated
with xA.
i. Error of xA: err (xa) = Xt-Xa
ii. Absolute error of xa: Abs(Xa) = |Xt - Xa|
iii. Relative error of xa: Rel(Xa) = |Xt- Xa|/|Xt|
iv. Number of significant cant digits
No comments:
Post a Comment