I want to ask a silly question. When using a normal distribution as likelihood function, fit a curve, the values on the curve is denoted as obs like this

```
with pm.Model() as background_model:
amp = pm.Uniform('amp',lower=-100,upper=-0)
rate = pm.Uniform('rate', lower=0,upper= 200)
t_start = pm.Uniform('t_start',lower=0,upper=500)
b1=pm.Uniform('b1',lower=0,upper=4)
y_observed=pm.Normal(
"y_observed",
mu=amp*np.exp(-(t_back-t_start)/rate)+b1,
sigma=noise,
observed=obs,
)
output = pm.Deterministic('output',amp*np.exp(-(t_back-t_start)/rate)+b1)
prior = pm.sample_prior_predictive()
posterior_b = pm.sample(draws = Samples, target_accept = 0.9,chains=4,cores=1)
posterior_only_background = pm.sample_posterior_predictive(posterior_b)
```

how do we determine the selection of the parameter sigma? What factors are usually related to it?