Advanced Marketing Modeling 2018

Getting started with Python

Download and install Python3.6 from https://www.python.org/downloads/. Make sure to include it to your PATH variable if you are using Windows (just tick the box during the installation).

To run the Python code from this repository you may need to install additional modules. You can do this by opening a terminal (CTRL-ALT-T on Linux or MacOS) or a command line (on Windows just push HOME and start typing run, you should see a black icon).

First you want to change into your directory:

cd __path_to_your_cloned_repository__

Next you want to install the packages from the requirements.txt-file:

pip install -r requirements.txt

Thats all. After that you can run the python scripts by running:

python __name_of_script__.py

Note that this might be different if you are using IDE's like PyCharm, IDLE, Spyder, etc.

Weather data

This script scrapes the web for weather data, interfacing with the Dark Sky API in order to obtain data about temperature and weather conditions measured at Boston Logan International Airport. At the end you will write a csv-file to the data directory of this repository (on your local disk, of course).

How to run this script

1. Get a key from https://darksky.net/dev (don't worry, it's for free).
2. install the darksky-module from github, this seems to be the only version that is not broken:

pip install git+https://github.com/zachwill/darksky.git --user

3. Save your key in the key.txt-file.
4. run the weather.py script.

Functions

ping_darksky(time, key)

Interface to Dark Sky API. Requests data from Boston Airport for a specific time.

• Arguments:

  • time: a datetime object. Denotes the day of the requested record.
  • key: a character string. Key to interface the Dark Sky API.

• Returns:

  • Dictionary containing:
    • day: datetime of the observation (in timestamp format).
    • tempMin: minimum temperature in degrees Fahrenheit at that date.
    • tempMax: maximum temperature in degrees Fahrenheit at that date.
    • summary: weather summary of that date.
    • desc: single word description of weather conditions.

switch_key():

Key generator that allows to switch between keys that are provided in the secret_key.txt file.

• Yields:
  • Key to access Dark Sky API.

models

Models for AMM 18.

Create handy class interface for estimation, calculation of elasticities, simulation of price changes and generation of Latex tables.

AMM

AMM(self)

Model class for the 2018 Advanced Marketing Modeling lecture.

Estimate a model, calculate elasticities and simulate the effects of a price change in just a few lines.

Examples:

model = "L5+C(year)+price_liter+C(display_all)+C(feature_all)"
modifiers = {"DIET COKE": 0.25, "COKE CLASSIC": 1.25}
amm = AMM()
amm.fit(data=data.loc[data.PACKAGE == "BOTTLE"],
        RHS=model,
        share="share",
        agg=["year", "week"])
amm.summary()
amm.simulate(modifiers)
amm.to_latex()

fit

AMM.fit(self, data, RHS, share, agg=None, nests=None, grp_unit='liters', price_var='price_liter')

Fit AMM-model to data and calculate elasticities.

The type of model is specified through the nests parameter. If nests is None (the default) an Aggregated Logit Model is fitted. If nests is a list, a Nested Aggregated Logit Model is fitted. The length of the list determines the nesting structure. A length of one corresponds to a 2 level Nested Aggregated Logit Model and a length of two corresponds to a 3 level Nested Aggregated Logit Model.

Note that data has very specific assumptions in place. It assumes an IRI format which means it expects following columns to exist:

• L5: brand names
• share: precomputed shares for each observation.
• grp_unit: precomputed measure of volume.
• price_var: precomputed measure of price.

Where L5 is necessary and each column can be customized to allow for different ways to infer the shares of the outside good (you don't need to calculate the outside good explicitly).

For example: it is possible to compute shares of each product based on:

• total_rev_cola
• total_rev_carbbev
• total_vol_cola
• total_vol_carbbev

Pick one, calculate the shares (by yourself), store the results as a column in data and set share to that column's name. Similarly, you need to proceed with grp_unit and price_var.

Parameters:

• data: a pandas.DataFrame formatted in the IRI format.
• RHS: string of variables used in the OLS model.
• share: column of data that stores the shares of the products.
• agg: list of most coarse aggregation level. Defaults to None
• nests: list of nesting structure.
• grp_unit: column of data that specifies group aggregation units.
• price_var: column of data that specifies prices to be used.

Returns:

• self

Examples:

model = "L5+C(year)+price_liter+C(display_all)+C(feature_all)"
amm = AMM()
amm.fit(data=data.loc[data.PACKAGE == "BOTTLE"],
        RHS=model,
        share="share",
        agg=["year", "week"])

elasticities

AMM.elasticities(self)

Calculate elasticities.

Returns

• table of estimated elasticities

Examples:

amm.elasticities()

simulate

AMM.simulate(self, modifiers={})

Simulate price changes.

Parameters:

• modifiers: dictionary
  • key: brand, must appear in data's L5 column
  • value: multiplier of price

Returns:

• self

Examples:

modifiers = {"DIET COKE": 0.25, "COKE CLASSIC": 1.25}
amm.simulate(modifiers)

summary

AMM.summary(self)

Print summary of OLS-model and estimated elasticities to stdout.

to_latex

AMM.to_latex(self, prefix='')

Write results as Latex-tables to disk.

Creates three files:

• ols_results.tex: file containing OLS-summary tables
• elasticities.tex: file containing elasticity-tables
• simulation.tex: file containing simulation-tables

Parameters:

• prefix: string to prepend filename (defaults to empty string)