---
myst:
    substitutions:
        include:
            python: "`numpy array`"
            cpp: "`vector<double>`"
---

(calibration_classes_functions_observations_calibration_classes_estimate_residuals)=
# Estimate custom residuals

Once the estimate has been performed, both *a priori* and *a posteriori* residuals can be computed. 
Since *a priori* and *a posteriori* estimates depend on the chosen algorithm, their formulas are not detailed 
here. It is also possible to re-evaluate residuals for any custom set of parameter values provided by the 
user. This can be done through

````{only} cpp
```cpp
void estimateCustomResiduals(string resName, int theta_nb, double *theta_val);
```
````

````{only} py
```python
estimateCustomResiduals(resName, theta_nb, theta_val)
```
````

It takes three arguments, which are:

1. **resName**: a tag identifying the stored set of residuals;

2. **theta_nb**: the number of parameter values in the array (must match `_nPar`);

3. **theta_val**: an array containing the parameter values.

For convenience, this method can be called with a {{include[language]}}, as shown below

````{only} cpp
```cpp
vector<double> mypar = {0., 2., 3.};
mycal->estimateCustomResiduals("set1", mypar.size(), &mypar[0]);
```
````

````{only} py
```python
import numpy as np
mypar = np.array([0., 2., 3.])
mycal.estimateCustomResiduals("set1", len(mpar), mypar)
```
````
This feature allows re-estimating the residuals based on the available sample after the *a posteriori* distributions 
have been calibrated. A common use case is Markov chain Monte Carlo methods, where it may be necessary to discard the 
burn-in phase before evaluating residuals. Once created, the custom set can be used in the `drawResiduals` function, 
as discussed in [](#calibration_classes_functions_observations_calibration_classes_draw_residuals).