2.4 Validation
Model verification and validation are the primary processes for quantifying and
building credibility in numerical models. Verification is the process of determining
that a model implementation accurately represents the developer’s conceptual
description of the model and its solution. Validation is the process of determining
the degree to which a model is an accurate representation of the real world from
the perspective of the intended uses of the model. Both verification and validation
are processes that accumulative evidence of a model’s correctness or accuracy for
a real situation; thus verification and validation cannot prove that a model is
correct and accurate for all possible scenarios, but, rather, it can provide evidence
that the model is sufficiently accurate for its intended use. It is important to
perform a proper validation and verification to the model to extrapolate the results
to other similar situations and predict new events.
The data used for model validation have not been used to build the model.
There is a strong reason for not using data as we used in parameter estimation, it
will make us think that the model gives better predictions than it is really capable.
In linear regression, we correct for the effect of estimating two parameters by
dividing the residual sum of squares by (n-2) instead of (n). Since data often
contain errors, it sometimes sufficient to prove that there is a not statistically
significant difference between experimental data and model predictions (Bender
1978).
The differences between data and model predictions can be measured by
graphical comparison, confidence intervals, and statistical tests. Hypothesis tests
are particularly useful in comparing distributions, variances, or time series of
model outputs to determine whether the model predictions are within an acceptable range of precision.
2.4.1 Validation Criteria
Some consist of comparisons with results available, others require the generation
of new results for comparison, while others are based on responses from experts.
Some of these criteria are used for verification to model validation. According to
Godoy and Dardati (2001), the validation types can be classified as follows:
Comparison with other numerical solutions: the results of the model are associated with another solution identifying the proximity between them, that is,
approximate results validate other approximate also obtained by other authors or
other methods.
This technique does not involve a check of the representation of reality model.
This type of comparison is also used to verify the model that can be verified when
it is otherwise (Freedman and Ibaraki 2003).
2 Mathematical Modeling of Biosystems
63
Model verification and validation are the primary processes for quantifying and
building credibility in numerical models. Verification is the process of determining
that a model implementation accurately represents the developer’s conceptual
description of the model and its solution. Validation is the process of determining
the degree to which a model is an accurate representation of the real world from
the perspective of the intended uses of the model. Both verification and validation
are processes that accumulative evidence of a model’s correctness or accuracy for
a real situation; thus verification and validation cannot prove that a model is
correct and accurate for all possible scenarios, but, rather, it can provide evidence
that the model is sufficiently accurate for its intended use. It is important to
perform a proper validation and verification to the model to extrapolate the results
to other similar situations and predict new events.
The data used for model validation have not been used to build the model.
There is a strong reason for not using data as we used in parameter estimation, it
will make us think that the model gives better predictions than it is really capable.
In linear regression, we correct for the effect of estimating two parameters by
dividing the residual sum of squares by (n-2) instead of (n). Since data often
contain errors, it sometimes sufficient to prove that there is a not statistically
significant difference between experimental data and model predictions (Bender
1978).
The differences between data and model predictions can be measured by
graphical comparison, confidence intervals, and statistical tests. Hypothesis tests
are particularly useful in comparing distributions, variances, or time series of
model outputs to determine whether the model predictions are within an acceptable range of precision.
2.4.1 Validation Criteria
Some consist of comparisons with results available, others require the generation
of new results for comparison, while others are based on responses from experts.
Some of these criteria are used for verification to model validation. According to
Godoy and Dardati (2001), the validation types can be classified as follows:
Comparison with other numerical solutions: the results of the model are associated with another solution identifying the proximity between them, that is,
approximate results validate other approximate also obtained by other authors or
other methods.
This technique does not involve a check of the representation of reality model.
This type of comparison is also used to verify the model that can be verified when
it is otherwise (Freedman and Ibaraki 2003).
2 Mathematical Modeling of Biosystems
63