Files
OpenBB/website/content/sdk/reference/qa/var.md
jose-donato 2e7eb4c9b6 New documentation website (#3526)
* fix(docs): added margin between sections

* feat: removed unused files; updated ga

* fix: juan stuff doesnt work

* fix: random fix

* moon at noon

* fix: update docusaurus config

* Update controller_doc_classes.py

* fix: temp fix

* fix: temp fix

* Update controller_doc_classes.py

* fix: remove old content

* fix: taggfix

* init

* fix: styling and moved file

* Update generate_terminal_markdown.py

* fix: removed unused images

* fix: port new files

* Update generate_sdk_markdown.py

* Update controller_doc_classes.py

Co-authored-by: teh_coderer <me@tehcoderer.com>
2022-11-22 16:47:03 +00:00

85 lines
2.5 KiB
Markdown
Vendored

---
title: var
description: OpenBB SDK Function
---
import Tabs from '@theme/Tabs';
import TabItem from '@theme/TabItem';
# var
<Tabs>
<TabItem value="model" label="Model" default>
Gets value at risk for specified stock dataframe.
Source Code: [[link](https://github.com/OpenBB-finance/OpenBBTerminal/tree/main/openbb_terminal/common/quantitative_analysis/qa_model.py#L224)]
```python
openbb.qa.var(data: pd.DataFrame, use_mean: bool = False, adjusted_var: bool = False, student_t: bool = False, percentile: Union[int, float] = 99.9, portfolio: bool = False)
```
---
## Parameters
| Name | Type | Description | Default | Optional |
| ---- | ---- | ----------- | ------- | -------- |
| data | pd.DataFrame | Data dataframe | None | False |
| use_mean | bool | If one should use the data mean for calculation | False | True |
| adjusted_var | bool | If one should return VaR adjusted for skew and kurtosis | False | True |
| student_t | bool | If one should use the student-t distribution | False | True |
| percentile | Union[int,float] | VaR percentile | 99.9 | True |
| portfolio | bool | If the data is a portfolio | False | True |
---
## Returns
| Type | Description |
| ---- | ----------- |
| pd.DataFrame | DataFrame with Value at Risk per percentile |
---
</TabItem>
<TabItem value="view" label="Chart">
Prints table showing VaR of dataframe.
Source Code: [[link](https://github.com/OpenBB-finance/OpenBBTerminal/tree/main/openbb_terminal/common/quantitative_analysis/qa_view.py#L1048)]
```python
openbb.qa.var_chart(data: pd.DataFrame, symbol: str = "", use_mean: bool = False, adjusted_var: bool = False, student_t: bool = False, percentile: float = 99.9, data_range: int = 0, portfolio: bool = False)
```
---
## Parameters
| Name | Type | Description | Default | Optional |
| ---- | ---- | ----------- | ------- | -------- |
| data | pd.Dataframe | Data dataframe | None | False |
| use_mean | bool | if one should use the data mean return | False | True |
| symbol | str | name of the data | | True |
| adjusted_var | bool | if one should have VaR adjusted for skew and kurtosis (Cornish-Fisher-Expansion) | False | True |
| student_t | bool | If one should use the student-t distribution | False | True |
| percentile | int | var percentile | 99.9 | True |
| data_range | int | Number of rows you want to use VaR over | 0 | True |
| portfolio | bool | If the data is a portfolio | False | True |
---
## Returns
This function does not return anything
---
</TabItem>
</Tabs>