| Name | localprojections JSON |
| Version |
0.1.5
JSON |
| download |
| home_page | None |
| Summary | This module implements the local projections models for single entity time series and panel / longitudinal data, as well as threshold versions. |
| upload_time | 2024-04-25 05:27:33 |
| maintainer | None |
| docs_url | None |
| author | None |
| requires_python | >=3.9 |
| license | MIT License Copyright (c) 2024 Jing Lian Suah Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. |
| keywords |
econometrics
local projections
panel data
time series
|
| VCS |
 |
| bugtrack_url |
|
| requirements |
No requirements were recorded.
|
| Travis-CI |
No Travis.
|
| coveralls test coverage |
No coveralls.
|
# localprojections
This module implements the local projections models for single entity time series, and panel / longitudinal data settings, due to Jorda (2005), and based on codes available [here](https://sites.google.com/site/oscarjorda/home/local-projections).
# Installation
1. ```pip install localprojections```
# Implementation
## Panel Local Projections Model
### Documentation
```python
localprojections.PanelLP(data, Y, response, horizon, lags, varcov, ci_width)
```
#### Parameters
data :
Pandas MultiIndex dataframe with entity as the outer index, and time as the inner index.
Y :
List of column names in ```data``` to be used in the model estimation
response :
List of column names in ```Y``` to be used as response variables when estimating the impulse response functions (IRFs)
horizon :
Integer indicating the estimation horizon of the IRFs
lags :
Integer indicating the number of lags to be included in the model estimation
varcov :
Variance-covariance estimator to be used in estimating standard errors; refer to the [linearmodels package](https://bashtage.github.io/linearmodels/panel/panel/linearmodels.panel.model.PanelOLS.fit.html#linearmodels.panel.model.PanelOLS.fit).
ci_width :
Float higher than 0 and less than 1, i.e., (0, 1), indicating the width of the confidence intervals of the IRFs; ```ci_width=0.95``` indicates a 95% confidence interval
#### Output
This function returns a pandas dataframe of 6 columns:
1. ```Shock``` indicates the shock variable
2. ```Response``` indicates the response variable
3. ```Horizon``` indicates the response horizon of the IRF
4. ```Mean``` indicates the point estimate of the IRF
5. ```LB``` indicates the lower bound of the confidence interval of the IRF
6. ```LB``` indicates the upper bound of the confidence interval of the IRF
For instance, the estimates of the 6-period ahead IRF of y from a shock in x, can be found in the row with ```Shock=x```, ```Response=y```, and ```Horizon=6```
### Example
```python
from statsmodels.datasets import grunfeld
import localprojections as lp
df = grunfeld.load_pandas().data # import the Grunfeld investment data set
df = df.set_index(['firm', 'year']) # set entity-year indices (as per requirements in bashtage's linearmodels)
endog = ['invest', 'value', 'capital'] # cholesky ordering: invest --> value --> capital
response = endog.copy() # estimate the responses of all variables to shocks from all variables
irf_horizon = 8 # estimate IRFs up to 8 periods ahead
opt_lags = 2 # include 2 lags in the local projections model
opt_cov = 'robust' # HAC standard errors
opt_ci = 0.95 # 95% confidence intervals
irf = lp.PanelLP(data=df, # input dataframe
Y=endog, # variables in the model
response=response, # variables whose IRFs should be estimated
horizon=irf_horizon, # estimation horizon of IRFs
lags=opt_lags, # lags in the model
varcov=opt_cov, # type of standard errors
ci_width=opt_ci # width of confidence band
)
irfplot = lp.IRFPlot(irf=irf, # take output from the estimated model
response=['invest'], # plot only response of invest ...
shock=endog, # ... to shocks from all variables
n_columns=2, # max 2 columns in the figure
n_rows=2, # max 2 rows in the figure
maintitle='Panel LP: IRFs of Investment', # self-defined title of the IRF plot
show_fig=True, # display figure (from plotly)
save_pic=False # don't save any figures on local drive
)
```
## Panel Local Projections Model with Exogenous Variables (Panel LPX)
### Documentation
```python
localprojections.PanelLPX(data, Y, X, response, horizon, lags, varcov, ci_width)
```
#### Parameters
data :
Pandas MultiIndex dataframe with entity as the outer index, and time as the inner index.
Y :
List of column names in ```data``` to be used in the model estimation as endogenous variables
X :
List of column names in ```data``` to be used in the model estimation as exogenous variables
response :
List of column names in ```Y``` to be used as response variables when estimating the impulse response functions (IRFs)
horizon :
Integer indicating the estimation horizon of the IRFs
lags :
Integer indicating the number of lags to be included in the model estimation
varcov :
Variance-covariance estimator to be used in estimating standard errors; refer to the [linearmodels package](https://bashtage.github.io/linearmodels/panel/panel/linearmodels.panel.model.PanelOLS.fit.html#linearmodels.panel.model.PanelOLS.fit).
ci_width :
Float higher than 0 and less than 1, i.e., (0, 1), indicating the width of the confidence intervals of the IRFs; ```ci_width=0.95``` indicates a 95% confidence interval
#### Output
This function returns a pandas dataframe of 6 columns:
1. ```Shock``` indicates the shock variable
2. ```Response``` indicates the response variable
3. ```Horizon``` indicates the response horizon of the IRF
4. ```Mean``` indicates the point estimate of the IRF
5. ```LB``` indicates the lower bound of the confidence interval of the IRF
6. ```LB``` indicates the upper bound of the confidence interval of the IRF
For instance, the estimates of the 6-period ahead IRF of y from a shock in x, can be found in the row with ```Shock=x```, ```Response=y```, and ```Horizon=6```
### Example
```python
from statsmodels.datasets import grunfeld
import localprojections as lp
df = grunfeld.load_pandas().data # import the Grunfeld investment data set
df = df.set_index(['firm', 'year']) # set entity-year indices (as per requirements in bashtage's linearmodels)
endog = ['invest', 'value', 'capital'] # cholesky ordering: invest --> value --> capital
response = endog.copy() # estimate the responses of all variables to shocks from all variables
irf_horizon = 8 # estimate IRFs up to 8 periods ahead
opt_lags = 2 # include 2 lags in the local projections model
opt_cov = 'robust' # HAC standard errors
opt_ci = 0.95 # 95% confidence intervals
irf = lp.PanelLP(data=df, # input dataframe
Y=endog, # variables in the model
response=response, # variables whose IRFs should be estimated
horizon=irf_horizon, # estimation horizon of IRFs
lags=opt_lags, # lags in the model
varcov=opt_cov, # type of standard errors
ci_width=opt_ci # width of confidence band
)
irfplot = lp.IRFPlot(irf=irf, # take output from the estimated model
response=['invest'], # plot only response of invest ...
shock=endog, # ... to shocks from all variables
n_columns=2, # max 2 columns in the figure
n_rows=2, # max 2 rows in the figure
maintitle='Panel LP: IRFs of Investment', # self-defined title of the IRF plot
show_fig=True, # display figure (from plotly)
save_pic=False # don't save any figures on local drive
)
```
## Panel Local Projections Model with Exogenous Variables (Panel LPX)
### Documentation
```python
localprojections.ThresholdPanelLPX(data, Y, X, threshold_var, response, horizon, lags, varcov, ci_width)
```
#### Parameters
data :
Pandas MultiIndex dataframe with entity as the outer index, and time as the inner index.
Y :
List of column names in ```data``` to be used in the model estimation as endogenous variables
X :
List of column names in ```data``` to be used in the model estimation as exogenous variablesiables
threshold_var :
String indicating column in ```data``` to be used as the threshold variable; must take values 0 or 1 for technically correct implementation
response :
List of column names in ```Y``` to be used as response variables when estimating the impulse response functions (IRFs)
horizon :
Integer indicating the estimation horizon of the IRFs
lags :
Integer indicating the number of lags to be included in the model estimation
varcov :
Variance-covariance estimator to be used in estimating standard errors; refer to the [linearmodels package](https://bashtage.github.io/linearmodels/panel/panel/linearmodels.panel.model.PanelOLS.fit.html#linearmodels.panel.model.PanelOLS.fit).
ci_width :
Float higher than 0 and less than 1, i.e., (0, 1), indicating the width of the confidence intervals of the IRFs; ```ci_width=0.95``` indicates a 95% confidence interval
#### Output
This function returns *two* pandas dataframes of 6 columns each, with the first output corresponding to when ```threshold_var``` takes value 1, and the second when ```threshold_var`` takes value 0:
1. ```Shock``` indicates the shock variable
2. ```Response``` indicates the response variable
3. ```Horizon``` indicates the response horizon of the IRF
4. ```Mean``` indicates the point estimate of the IRF
5. ```LB``` indicates the lower bound of the confidence interval of the IRF
6. ```LB``` indicates the upper bound of the confidence interval of the IRF
For instance, the estimates of the 6-period ahead IRF of y from a shock in x, can be found in the row with ```Shock=x```, ```Response=y```, and ```Horizon=6```
## Single Entity Time Series Local Projections Model
### Documentation
```python
localprojections.TimeSeriesLP(data, Y, response, horizon, lags, newey_lags, ci_width)
```
#### Parameters
data :
Pandas dataframe
Y :
List of column names in ```data``` to be used in the model estimation
response :
List of column names in ```Y``` to be used as response variables when estimating the impulse response functions (IRFs)
horizon :
Integer indicating the estimation horizon of the IRFs
lags :
Integer indicating the number of lags to be included in the model estimation
newey_lags :
Maximum number of lags to be used when estimating the Newey-West standard errors
ci_width :
Float higher than 0 and less than 1, i.e., (0, 1), indicating the width of the confidence intervals of the IRFs; ```ci_width=0.95``` indicates a 95% confidence interval
#### Output
This function also returns a pandas dataframe of 6 columns:
1. ```Shock``` indicates the shock variable
2. ```Response``` indicates the response variable
3. ```Horizon``` indicates the response horizon of the IRF
4. ```Mean``` indicates the point estimate of the IRF
5. ```LB``` indicates the lower bound of the confidence interval of the IRF
6. ```LB``` indicates the upper bound of the confidence interval of the IRF
For instance, the estimates of the 6-period ahead IRF of y from a shock in x, can be found in the row with ```Shock=x```, ```Response=y```, and ```Horizon=6```
### Example
```python
from statsmodels.datasets import grunfeld
import localprojections as lp
df = grunfeld.load_pandas().data # import the Grunfeld investment data set
df = df[df['firm'] == 'General Motors'] # keep only one entity (as an example of a single entity time series setting)
df = df.set_index(['year']) # set time variable as index
endog = ['invest', 'value', 'capital'] # cholesky ordering: invest --> value --> capital
response = endog.copy() # estimate the responses of all variables to shocks from all variables
irf_horizon = 8 # estimate IRFs up to 8 periods ahead
opt_lags = 2 # include 2 lags in the local projections model
opt_cov = 'robust' # HAC standard errors
opt_ci = 0.95 # 95% confidence intervals
# Use TimeSeriesLP for the single entity case
irf = lp.TimeSeriesLP(data=df, # input dataframe
Y=endog, # variables in the model
response=response, # variables whose IRFs should be estimated
horizon=irf_horizon, # estimation horizon of IRFs
lags=opt_lags, # lags in the model
newey_lags=2, # maximum lags when estimating Newey-West standard errors
ci_width=opt_ci # width of confidence band
)
irfplot = lp.IRFPlot(irf=irf, # take output from the estimated model
response=['invest'], # plot only response of invest ...
shock=endog, # ... to shocks from all variables
n_columns=2, # max 2 columns in the figure
n_rows=2, # max 2 rows in the figure
maintitle='Single Entity Time Series LP: IRFs of Investment', # self-defined title of the IRF plot
show_fig=True, # display figure (from plotly)
save_pic=False # don't save any figures on local drive
)
```
## Plotting Impulse Response Functions
### Documentation
```python
localprojections.IRFPlot(irf, response, shock, n_columns, n_rows, maintitle, show_fig, save_pic, out_path, out_name, annot_size, font_size)
```
#### Parameters
irf :
Output from ```PanelLP()```, or ```TimeSeriesLP()```
response :
List of variables contained in ```irf```'s ```Response``` column whose IRFs is to be plotted
shock :
List of variables contained in ```irf```'s ```Shock``` column whose IRFs is to be plotted
n_columns :
Integer indicating the number of IRF figures per row in the overall figure
n_rows :
Integer indicating the number of IRF figures per column in the overall figure
maintitle :
Strings to be used as the title of the overall figure; default is ```''Local Projections Model: Impulse Response Functions'```
show_fig :
Boolean indicating whether to render the overall figure
save_pic :
Boolean indicating whether to save the overall figure in the local directory; if ```True```, a ```html``` file and a ```png``` file will be saved
out_path :
Strings indicating the directory at which the overall figure should be saved in; only used if ```save_pic``` is ```True```
out_name :
Strings indicating the name of the file in which the overall figure should be saved as; only used if ```save_pic``` is ```True```, and default is ```IRFPlot```
annot_size :
Integer indicating the font size of titles of each subplot in the figure; defaults to 6
font_size :
Integer indicating the font size of the title, and axes labels; defaults to 9
#### Output
This function returns a [plotly graph objects figure](https://plotly.com/python-api-reference/generated/plotly.graph_objects.Figure.html) with ```n_columns``` (columns) x ```n_rows``` (rows) subplots. Depending on arguments passed, the figure may be rendered during implementation and / or saved in the local directory.
### Examples
See above.
# Requirements
## Python Packages
- pandas>=1.4.3
- numpy>=1.23.0
- linearmodels>=4.27
- plotly>=5.9.0
- statsmodels>=0.13.2
Raw data
{
"_id": null,
"home_page": null,
"name": "localprojections",
"maintainer": null,
"docs_url": null,
"requires_python": ">=3.9",
"maintainer_email": null,
"keywords": "Econometrics, Local Projections, Panel Data, Time Series",
"author": null,
"author_email": "Jing Lian Suah <suahjinglian@bnm.gov.my>",
"download_url": "https://files.pythonhosted.org/packages/b7/3a/3b848702814a73f3db7a990d6a6c1b8d8a668c313c5b78bd0a77f5d6ca65/localprojections-0.1.5.tar.gz",
"platform": null,
"description": "# localprojections\r\nThis module implements the local projections models for single entity time series, and panel / longitudinal data settings, due to Jorda (2005), and based on codes available [here](https://sites.google.com/site/oscarjorda/home/local-projections).\r\n\r\n# Installation\r\n1. ```pip install localprojections```\r\n\r\n# Implementation\r\n## Panel Local Projections Model\r\n### Documentation\r\n```python\r\nlocalprojections.PanelLP(data, Y, response, horizon, lags, varcov, ci_width)\r\n```\r\n#### Parameters\r\ndata : \r\n\tPandas MultiIndex dataframe with entity as the outer index, and time as the inner index.\r\n\r\nY : \r\n\tList of column names in ```data``` to be used in the model estimation\r\n\r\nresponse : \r\n\tList of column names in ```Y``` to be used as response variables when estimating the impulse response functions (IRFs)\r\n\r\nhorizon : \r\n\tInteger indicating the estimation horizon of the IRFs\r\n\r\nlags : \r\n\tInteger indicating the number of lags to be included in the model estimation\r\n\r\nvarcov : \r\n\tVariance-covariance estimator to be used in estimating standard errors; refer to the [linearmodels package](https://bashtage.github.io/linearmodels/panel/panel/linearmodels.panel.model.PanelOLS.fit.html#linearmodels.panel.model.PanelOLS.fit).\r\n\r\nci_width : \r\n\tFloat higher than 0 and less than 1, i.e., (0, 1), indicating the width of the confidence intervals of the IRFs; ```ci_width=0.95``` indicates a 95% confidence interval\r\n\r\n#### Output\r\nThis function returns a pandas dataframe of 6 columns: \r\n1. ```Shock``` indicates the shock variable\r\n2. ```Response``` indicates the response variable\r\n3. ```Horizon``` indicates the response horizon of the IRF\r\n4. ```Mean``` indicates the point estimate of the IRF\r\n5. ```LB``` indicates the lower bound of the confidence interval of the IRF\r\n6. ```LB``` indicates the upper bound of the confidence interval of the IRF\r\n\r\nFor instance, the estimates of the 6-period ahead IRF of y from a shock in x, can be found in the row with ```Shock=x```, ```Response=y```, and ```Horizon=6```\r\n\r\n### Example\r\n\r\n```python\r\nfrom statsmodels.datasets import grunfeld\r\nimport localprojections as lp\r\n\r\ndf = grunfeld.load_pandas().data # import the Grunfeld investment data set\r\ndf = df.set_index(['firm', 'year']) # set entity-year indices (as per requirements in bashtage's linearmodels)\r\n\r\nendog = ['invest', 'value', 'capital'] # cholesky ordering: invest --> value --> capital\r\nresponse = endog.copy() # estimate the responses of all variables to shocks from all variables\r\nirf_horizon = 8 # estimate IRFs up to 8 periods ahead\r\nopt_lags = 2 # include 2 lags in the local projections model\r\nopt_cov = 'robust' # HAC standard errors\r\nopt_ci = 0.95 # 95% confidence intervals\r\n\r\nirf = lp.PanelLP(data=df, # input dataframe\r\n Y=endog, # variables in the model\r\n response=response, # variables whose IRFs should be estimated\r\n horizon=irf_horizon, # estimation horizon of IRFs\r\n lags=opt_lags, # lags in the model\r\n varcov=opt_cov, # type of standard errors\r\n ci_width=opt_ci # width of confidence band\r\n )\r\nirfplot = lp.IRFPlot(irf=irf, # take output from the estimated model\r\n response=['invest'], # plot only response of invest ...\r\n shock=endog, # ... to shocks from all variables\r\n n_columns=2, # max 2 columns in the figure\r\n n_rows=2, # max 2 rows in the figure\r\n maintitle='Panel LP: IRFs of Investment', # self-defined title of the IRF plot\r\n show_fig=True, # display figure (from plotly)\r\n save_pic=False # don't save any figures on local drive\r\n )\r\n```\r\n\r\n## Panel Local Projections Model with Exogenous Variables (Panel LPX)\r\n### Documentation\r\n```python\r\nlocalprojections.PanelLPX(data, Y, X, response, horizon, lags, varcov, ci_width)\r\n```\r\n#### Parameters\r\ndata : \r\n\tPandas MultiIndex dataframe with entity as the outer index, and time as the inner index.\r\n\r\nY : \r\n\tList of column names in ```data``` to be used in the model estimation as endogenous variables\r\n\r\nX : \r\n\tList of column names in ```data``` to be used in the model estimation as exogenous variables\r\n\r\nresponse : \r\n\tList of column names in ```Y``` to be used as response variables when estimating the impulse response functions (IRFs)\r\n\r\nhorizon : \r\n\tInteger indicating the estimation horizon of the IRFs\r\n\r\nlags : \r\n\tInteger indicating the number of lags to be included in the model estimation\r\n\r\nvarcov : \r\n\tVariance-covariance estimator to be used in estimating standard errors; refer to the [linearmodels package](https://bashtage.github.io/linearmodels/panel/panel/linearmodels.panel.model.PanelOLS.fit.html#linearmodels.panel.model.PanelOLS.fit).\r\n\r\nci_width : \r\n\tFloat higher than 0 and less than 1, i.e., (0, 1), indicating the width of the confidence intervals of the IRFs; ```ci_width=0.95``` indicates a 95% confidence interval\r\n\r\n#### Output\r\nThis function returns a pandas dataframe of 6 columns: \r\n1. ```Shock``` indicates the shock variable\r\n2. ```Response``` indicates the response variable\r\n3. ```Horizon``` indicates the response horizon of the IRF\r\n4. ```Mean``` indicates the point estimate of the IRF\r\n5. ```LB``` indicates the lower bound of the confidence interval of the IRF\r\n6. ```LB``` indicates the upper bound of the confidence interval of the IRF\r\n\r\nFor instance, the estimates of the 6-period ahead IRF of y from a shock in x, can be found in the row with ```Shock=x```, ```Response=y```, and ```Horizon=6```\r\n\r\n### Example\r\n\r\n```python\r\nfrom statsmodels.datasets import grunfeld\r\nimport localprojections as lp\r\n\r\ndf = grunfeld.load_pandas().data # import the Grunfeld investment data set\r\ndf = df.set_index(['firm', 'year']) # set entity-year indices (as per requirements in bashtage's linearmodels)\r\n\r\nendog = ['invest', 'value', 'capital'] # cholesky ordering: invest --> value --> capital\r\nresponse = endog.copy() # estimate the responses of all variables to shocks from all variables\r\nirf_horizon = 8 # estimate IRFs up to 8 periods ahead\r\nopt_lags = 2 # include 2 lags in the local projections model\r\nopt_cov = 'robust' # HAC standard errors\r\nopt_ci = 0.95 # 95% confidence intervals\r\n\r\nirf = lp.PanelLP(data=df, # input dataframe\r\n Y=endog, # variables in the model\r\n response=response, # variables whose IRFs should be estimated\r\n horizon=irf_horizon, # estimation horizon of IRFs\r\n lags=opt_lags, # lags in the model\r\n varcov=opt_cov, # type of standard errors\r\n ci_width=opt_ci # width of confidence band\r\n )\r\nirfplot = lp.IRFPlot(irf=irf, # take output from the estimated model\r\n response=['invest'], # plot only response of invest ...\r\n shock=endog, # ... to shocks from all variables\r\n n_columns=2, # max 2 columns in the figure\r\n n_rows=2, # max 2 rows in the figure\r\n maintitle='Panel LP: IRFs of Investment', # self-defined title of the IRF plot\r\n show_fig=True, # display figure (from plotly)\r\n save_pic=False # don't save any figures on local drive\r\n )\r\n```\r\n\r\n## Panel Local Projections Model with Exogenous Variables (Panel LPX)\r\n### Documentation\r\n```python\r\nlocalprojections.ThresholdPanelLPX(data, Y, X, threshold_var, response, horizon, lags, varcov, ci_width)\r\n```\r\n#### Parameters\r\ndata : \r\n\tPandas MultiIndex dataframe with entity as the outer index, and time as the inner index.\r\n\r\nY : \r\n\tList of column names in ```data``` to be used in the model estimation as endogenous variables\r\n\r\nX : \r\n\tList of column names in ```data``` to be used in the model estimation as exogenous variablesiables\r\n\r\nthreshold_var : \r\n\tString indicating column in ```data``` to be used as the threshold variable; must take values 0 or 1 for technically correct implementation\r\n\r\nresponse : \r\n\tList of column names in ```Y``` to be used as response variables when estimating the impulse response functions (IRFs)\r\n\r\nhorizon : \r\n\tInteger indicating the estimation horizon of the IRFs\r\n\r\nlags : \r\n\tInteger indicating the number of lags to be included in the model estimation\r\n\r\nvarcov : \r\n\tVariance-covariance estimator to be used in estimating standard errors; refer to the [linearmodels package](https://bashtage.github.io/linearmodels/panel/panel/linearmodels.panel.model.PanelOLS.fit.html#linearmodels.panel.model.PanelOLS.fit).\r\n\r\nci_width : \r\n\tFloat higher than 0 and less than 1, i.e., (0, 1), indicating the width of the confidence intervals of the IRFs; ```ci_width=0.95``` indicates a 95% confidence interval\r\n\r\n#### Output\r\nThis function returns *two* pandas dataframes of 6 columns each, with the first output corresponding to when ```threshold_var``` takes value 1, and the second when ```threshold_var`` takes value 0: \r\n1. ```Shock``` indicates the shock variable\r\n2. ```Response``` indicates the response variable\r\n3. ```Horizon``` indicates the response horizon of the IRF\r\n4. ```Mean``` indicates the point estimate of the IRF\r\n5. ```LB``` indicates the lower bound of the confidence interval of the IRF\r\n6. ```LB``` indicates the upper bound of the confidence interval of the IRF\r\n\r\nFor instance, the estimates of the 6-period ahead IRF of y from a shock in x, can be found in the row with ```Shock=x```, ```Response=y```, and ```Horizon=6```\r\n\r\n## Single Entity Time Series Local Projections Model\r\n### Documentation\r\n```python\r\nlocalprojections.TimeSeriesLP(data, Y, response, horizon, lags, newey_lags, ci_width)\r\n```\r\n#### Parameters\r\ndata : \r\n\tPandas dataframe\r\n\r\nY : \r\n\tList of column names in ```data``` to be used in the model estimation\r\n\r\nresponse : \r\n\tList of column names in ```Y``` to be used as response variables when estimating the impulse response functions (IRFs)\r\n\r\nhorizon : \r\n\tInteger indicating the estimation horizon of the IRFs\r\n\r\nlags : \r\n\tInteger indicating the number of lags to be included in the model estimation\r\n\r\nnewey_lags : \r\n\tMaximum number of lags to be used when estimating the Newey-West standard errors\r\n\r\nci_width : \r\n\tFloat higher than 0 and less than 1, i.e., (0, 1), indicating the width of the confidence intervals of the IRFs; ```ci_width=0.95``` indicates a 95% confidence interval\r\n\r\n#### Output\r\nThis function also returns a pandas dataframe of 6 columns: \r\n1. ```Shock``` indicates the shock variable\r\n2. ```Response``` indicates the response variable\r\n3. ```Horizon``` indicates the response horizon of the IRF\r\n4. ```Mean``` indicates the point estimate of the IRF\r\n5. ```LB``` indicates the lower bound of the confidence interval of the IRF\r\n6. ```LB``` indicates the upper bound of the confidence interval of the IRF\r\n\r\nFor instance, the estimates of the 6-period ahead IRF of y from a shock in x, can be found in the row with ```Shock=x```, ```Response=y```, and ```Horizon=6```\r\n\r\n### Example\r\n```python\r\nfrom statsmodels.datasets import grunfeld\r\nimport localprojections as lp\r\n\r\ndf = grunfeld.load_pandas().data # import the Grunfeld investment data set\r\ndf = df[df['firm'] == 'General Motors'] # keep only one entity (as an example of a single entity time series setting)\r\ndf = df.set_index(['year']) # set time variable as index\r\n\r\nendog = ['invest', 'value', 'capital'] # cholesky ordering: invest --> value --> capital\r\nresponse = endog.copy() # estimate the responses of all variables to shocks from all variables\r\nirf_horizon = 8 # estimate IRFs up to 8 periods ahead\r\nopt_lags = 2 # include 2 lags in the local projections model\r\nopt_cov = 'robust' # HAC standard errors\r\nopt_ci = 0.95 # 95% confidence intervals\r\n\r\n# Use TimeSeriesLP for the single entity case\r\nirf = lp.TimeSeriesLP(data=df, # input dataframe\r\n Y=endog, # variables in the model\r\n response=response, # variables whose IRFs should be estimated\r\n horizon=irf_horizon, # estimation horizon of IRFs\r\n lags=opt_lags, # lags in the model\r\n newey_lags=2, # maximum lags when estimating Newey-West standard errors\r\n ci_width=opt_ci # width of confidence band\r\n )\r\nirfplot = lp.IRFPlot(irf=irf, # take output from the estimated model\r\n response=['invest'], # plot only response of invest ...\r\n shock=endog, # ... to shocks from all variables\r\n n_columns=2, # max 2 columns in the figure\r\n n_rows=2, # max 2 rows in the figure\r\n maintitle='Single Entity Time Series LP: IRFs of Investment', # self-defined title of the IRF plot\r\n show_fig=True, # display figure (from plotly)\r\n save_pic=False # don't save any figures on local drive\r\n )\r\n\r\n```\r\n\r\n## Plotting Impulse Response Functions\r\n### Documentation\r\n```python\r\nlocalprojections.IRFPlot(irf, response, shock, n_columns, n_rows, maintitle, show_fig, save_pic, out_path, out_name, annot_size, font_size)\r\n```\r\n#### Parameters\r\nirf : \r\n\tOutput from ```PanelLP()```, or ```TimeSeriesLP()```\r\n\r\nresponse : \r\n\tList of variables contained in ```irf```'s ```Response``` column whose IRFs is to be plotted \r\n\r\nshock : \r\n\tList of variables contained in ```irf```'s ```Shock``` column whose IRFs is to be plotted \r\n\r\nn_columns : \r\n\tInteger indicating the number of IRF figures per row in the overall figure\r\n\r\nn_rows : \r\n\tInteger indicating the number of IRF figures per column in the overall figure\r\n\r\nmaintitle : \r\n\tStrings to be used as the title of the overall figure; default is ```''Local Projections Model: Impulse Response Functions'```\r\n\r\nshow_fig : \r\n\tBoolean indicating whether to render the overall figure\r\n\r\nsave_pic : \r\n\tBoolean indicating whether to save the overall figure in the local directory; if ```True```, a ```html``` file and a ```png``` file will be saved\r\n\r\nout_path : \r\n\tStrings indicating the directory at which the overall figure should be saved in; only used if ```save_pic``` is ```True```\r\n\r\nout_name : \r\n\tStrings indicating the name of the file in which the overall figure should be saved as; only used if ```save_pic``` is ```True```, and default is ```IRFPlot```\r\n\r\nannot_size : \r\n Integer indicating the font size of titles of each subplot in the figure; defaults to 6\r\n\r\nfont_size : \r\n Integer indicating the font size of the title, and axes labels; defaults to 9\r\n\r\n\r\n#### Output\r\nThis function returns a [plotly graph objects figure](https://plotly.com/python-api-reference/generated/plotly.graph_objects.Figure.html) with ```n_columns``` (columns) x ```n_rows``` (rows) subplots. Depending on arguments passed, the figure may be rendered during implementation and / or saved in the local directory.\r\n\r\n### Examples\r\nSee above.\r\n\r\n# Requirements\r\n## Python Packages\r\n- pandas>=1.4.3\r\n- numpy>=1.23.0\r\n- linearmodels>=4.27\r\n- plotly>=5.9.0\r\n- statsmodels>=0.13.2\r\n\r\n",
"bugtrack_url": null,
"license": "MIT License Copyright (c) 2024 Jing Lian Suah Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the \"Software\"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.",
"summary": "This module implements the local projections models for single entity time series and panel / longitudinal data, as well as threshold versions.",
"version": "0.1.5",
"project_urls": {
"GitHub": "https://github.com/suahjl/localprojections/",
"PyPI": "https://pypi.org/project/localprojections/"
},
"split_keywords": [
"econometrics",
" local projections",
" panel data",
" time series"
],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "28cdaf3e48be0d4a9e25ba6577395844e863cee489f9acb5a0ccc9fc8a1d4849",
"md5": "d374a8c6d28c8834fa1e92c52a27f558",
"sha256": "e7be7ce7b6f272d6a4b41901fb51cf37ee73661a166cf089b43ed32e50c67d85"
},
"downloads": -1,
"filename": "localprojections-0.1.5-py3-none-any.whl",
"has_sig": false,
"md5_digest": "d374a8c6d28c8834fa1e92c52a27f558",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": ">=3.9",
"size": 12458,
"upload_time": "2024-04-25T05:27:31",
"upload_time_iso_8601": "2024-04-25T05:27:31.001350Z",
"url": "https://files.pythonhosted.org/packages/28/cd/af3e48be0d4a9e25ba6577395844e863cee489f9acb5a0ccc9fc8a1d4849/localprojections-0.1.5-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "b73a3b848702814a73f3db7a990d6a6c1b8d8a668c313c5b78bd0a77f5d6ca65",
"md5": "7d9dae81cbfc42ded33e5b74c28ffc54",
"sha256": "0356727802483ebfd61dbf94f1ffb3c38aa4d65904dac2a76c923be7697555c7"
},
"downloads": -1,
"filename": "localprojections-0.1.5.tar.gz",
"has_sig": false,
"md5_digest": "7d9dae81cbfc42ded33e5b74c28ffc54",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.9",
"size": 14417,
"upload_time": "2024-04-25T05:27:33",
"upload_time_iso_8601": "2024-04-25T05:27:33.999935Z",
"url": "https://files.pythonhosted.org/packages/b7/3a/3b848702814a73f3db7a990d6a6c1b8d8a668c313c5b78bd0a77f5d6ca65/localprojections-0.1.5.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2024-04-25 05:27:33",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "suahjl",
"github_project": "localprojections",
"travis_ci": false,
"coveralls": false,
"github_actions": false,
"lcname": "localprojections"
}
