# Working with results

`fit()` returns a {py:class}`~pykinbiont.GrowthFitResults` which contains:

- `data` — the post-preprocessing `GrowthData` (smoothed, blank-subtracted, etc.)
- `results` — a list of {py:class}`~pykinbiont.CurveFitResult`, one per curve

## GrowthFitResults

### to_dataframe()

Returns a pandas DataFrame with one row per curve:

```python
df = results.to_dataframe()
print(df.columns.tolist())
# ['label', 'cluster', 'best_model', 'aic', 'loss', 'param_1', 'param_2', ...]
```

| Column | Description |
|---|---|
| `label` | Curve identifier |
| `cluster` | Cluster ID (integer, or `None` if no clustering was run) |
| `best_model` | Name of the winning model |
| `aic` | AICc of the best model |
| `loss` | Final loss value |
| `param_1`, `param_2`, … | Best-fit parameters in positional order |

### Iteration

`GrowthFitResults` is iterable and indexable:

```python
# Iterate
for r in results:
    print(r.label, r.best_model)

# Index
first = results[0]

# Length
print(len(results))
```

## CurveFitResult

Each `CurveFitResult` has:

| Attribute | Type | Description |
|---|---|---|
| `label` | `str` | Curve identifier |
| `best_model` | `str` | Name of the selected model |
| `best_params` | `list[float]` | Best-fit parameter values |
| `param_names` | `list[str]` | Parameter names (matches `best_params`) |
| `best_aic` | `float` | AICc of the best model |
| `fitted_curve` | `ndarray` | Fitted OD values (same time axis as `times`) |
| `times` | `ndarray` | Time axis for the fitted curve |
| `loss` | `float` | Final loss value |
| `all_results` | `list[dict]` | All candidate models with their params and AICc |

### Accessing parameters by name

```python
for r in results:
    params = dict(zip(r.param_names, r.best_params))
    print(f"{r.label}: {params}")
```

### Plotting fitted vs observed

```python
import matplotlib.pyplot as plt

fig, axes = plt.subplots(1, len(results), figsize=(4 * len(results), 3))

for ax, r in zip(axes, results):
    idx = results.data.labels.index(r.label)
    obs = results.data.curves[idx]

    ax.plot(results.data.times, obs, ".", alpha=0.4, label="data")
    ax.plot(r.times, r.fitted_curve, "-", label=r.best_model)
    ax.set_title(r.label)
    ax.set_xlabel("Time")
    ax.set_ylabel("OD")
    ax.legend(fontsize=7)

plt.tight_layout()
```

### Comparing candidate models

`all_results` contains every model that was tried:

```python
r = results[0]
for candidate in sorted(r.all_results, key=lambda x: x["aic"]):
    print(f"  {candidate['model_name']:20s}  AIC={candidate['aic']:.1f}  loss={candidate['loss']:.4f}")
```