Mean relative error

From DEBwiki
Jump to: navigation, search

AmP estimation

Data and completeness
Parameter estimation
Goodness-of-fit: SMSE / MRE
AmP Literature

Practice - essentials

Starting an estimation for a new species
Setting initial parameter values
Setting weight coefficients
Computing implied properties
Submitting to the collection
Obtaining parameter confidence intervals

Practice - extra modules
Code specification
User-defined files: run, mydata, pars_init, predict
Data: Zero-variate, Univariate, Pseudo-data

Typified models
Estimation options

For each entry a Mean Relative Error (MRE) is computed to help judge the goodness of fit.


To compute the MRE we start by computing the Relative Error (RE) for each data set. The RE of a data set is computed in this way

RE.png for SumWeights.png

where i refers to the data set and j to a given point in data set i. dij stands for the data, pij for the model prediction and wij for the associated weight coefficient. Finally, di in the denominator represents the average of all data points in set i (dij):

Mean d.png

In the case the sum of the weights wi is equal to zero we simply take the relative error to be zero. The implication is that this data set does not contribute to the MRE.

MRE is then the mean of all REi with wi greater than 0.

MRE RE.png

where n' is the number of data sets with wi greater than 0.

The zerovariate case

In the case of a zero-variate data set the relative error simplifies to


and the connection with the formal definition of relative is more easily seen.