Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
Adriano
@rsvp
Hello world!
Adriano
@rsvp
Added quandl module in dev branch for COTR position reports.
Adriano
@rsvp
Analysis of CFTC COTR reports using pandas -- committed to master yesterday, includes scale-free measures for precious metals, USD, bonds, and equities.
pandas can now easily read and analyze SEC 13F -- examples examine Druckenmiller and Paulson portfolios -- see SEC-13F-parse.ipynb
Adriano
@rsvp
BTW. the main repo is here: https://github.com/rsvp/fecon235 -- includes time-series data from the Federal Reserve.
Adriano
@rsvp
v2.15.1102 introduces two new notebooks: first examines gold contango. The second forecasts the Fed Funds rate one-year out using expected changes in LIBOR: https://github.com/rsvp/fecon235/blob/master/nb/qdl-libor-fed-funds.ipynb
Adriano
@rsvp
Shortened to https://git.io/fedfunds -- currently forecasts Fed funds effective rate at 50 bp for Dec 2016, so one rate hike then trade in upper target range, OR two rate hikes within one year then trade in lower target range.
Adriano
@rsvp
FOMC: Two more hours until Yellen speaks... in the meantime, check out our new Fed score: https://git.io/fed which is also useful to visualize the Phillips curve via sequential scatter plot with color heat map.
Adriano
@rsvp
As of v4.15.1230 all modules are Python 2 and 3 compatible. Import and preamble in notebooks have new cross-platform formats: please see https://git.io/fecon-intro for details. Bug report #2 filed for updating pre-2016 notebooks. -- Other than that: HAPPY NEW YEAR!
Adriano
@rsvp
Here's another illustration of "Fix issue #2 with v4 and p6 upgrades": https://git.io/georet which concerns comparative geometric mean returns.
Adriano
@rsvp
Fixed issue #2 for https://git.io/fedfunds -- NEW forecast for Fed Funds one-year out: 0.76% . Given the 2015-12-16 rate hike, this implies in terms of expected Fed action: one further rate hike of 25 bp (keeping actual in the upper range), or two rate hikes of 25 bp (keeping actual in the lower [0.75, 1.00] range). The findings from the notebook is now distilled into a NEW single convenient function forefunds().
Adriano
@rsvp
v4.16.0123 has newly revised position indicators for CFTC Commitment of Traders Report -- see https://git.io/cotr esp. for the Bonds/Equities joint narrative using color heat map. Normalized indicator for Equities has exceeded two standard deviations on the bearish side (for futures and options position, not relative to the price action).
Adriano
@rsvp
Using forefunds(), the forecast for Fed Funds one-year out: 0.55% which means NO hikes in the meanwhile (keeping actual at upper range), or slight chance of maybe one further rate hike (keeping actual at lower bound):
forefunds_2016-02-04
Adriano
@rsvp
BTW, we used the March cycle for the previous chart, i.e. forefunds('16h', '17h') -- the market's reversal of the 2015-12-16 rate hike is quite dramatic, since the Fed implied three more 25 bp rate hikes in their blue dot plot.
Adriano
@rsvp
Re: Housing economy https://git.io/housing completely revises hscore which roughly expresses the number of new "affordable" housing units in terms of percentage of total population. It confirms we are midway between the last housing bubble and the subprime mortgage crisis (leading to the Great Recession).
Adriano
@rsvp
Notebook on gold versus real rates updated: https://git.io/gold -- conjecture that real gold is a stationary time-series bound by real interest rates.
Adriano
@rsvp
Parsing SEC 13F filings, https://git.io/13F updated: we estimate that Paulson liquidated the equivalent of about 333,416 troy ounces at a weighted average price of $ 1059 at the end of 2015 -- that 's represents a -37.6% reduction in his GLD holdings, a gold ETF.
Adriano
@rsvp
Thanks to @samfisher83 for submitting our repo to ycombinator Hacker News: https://news.ycombinator.com/item?id=11240344
Adriano
@rsvp
NEW: on separable components of total return for equities, using Shiller's S&P stock market data, dating back to the year 1871: http://nbviewer.jupyter.org/github/rsvp/fecon235/blob/master/nb/qdl-spx-earn-div.ipynb
Adriano
@rsvp
Problem solely at GitHub, "Math Processing Error": rsvp/fecon235#3
Adriano
@rsvp
Cross-posted issue: GitHub LaTeX "Math Processing Error" when rendering Jupyter notebooks jupyter/help#21
Adriano
@rsvp
fecon235 now appears in a Chinese directory of Jupyter projects: http://docklet.unias.org/docklet-book/userguide/_book/zh/notebook/gallery.html -- so how to say thanks...
mmeanwe
@mmeanwe
what FRED tag/ID's are you using for fred-debt-pop.ipynb - specifically m4debt? Thx
Adriano
@rsvp
@gitmmeanwell hi, m4debt is pandas resampled monthly based on quarterly q4debt, whose fredcode is 'GFDEBTN', non-seasonally adjusted U.S. Federal debt in millions USD. The details are in the lib/yi_fred module, including the resampler monthly() which essentially does the linear interpolation.
mmeanwe
@mmeanwe
Got it - thx!
Adriano
@rsvp
A fecon235 user looks at trading algo module zipline: http://conrad.pythonanywhere.com/pydata/QuantFinance.html
Adriano
@rsvp
v4.16.0525 features new optimization module. Also, we finished up employment data, focusing on forecasting NFP, Nonfarm Payroll -- see http://nbviewer.jupyter.org/github/rsvp/fecon235/blob/master/nb/fred-employ-nfp.ipynb for details.
Adriano
@rsvp
Just based on the most recent NFP Nonfarm Payroll number, the market is no longer expecting the Fed to hike in July. The error bounds are important as detailed in our notebook... as it seems that the Federal Reserve is becoming data-point dependent for its monetary policies.
Adriano
@rsvp
Wondering if there was interest in running fecon235 via Docker container? Probably the quickest way for demos. Your thoughts?
Adriano
@rsvp
We set up the public repo for the fecon235 Docker image at https://hub.docker.com/r/rsvp/fecon235 -- just pull rsvp/fecon235 and then run this command, $ docker run -p 8888:8888 -it rsvp/fecon235 -- very simple, plus all the complicated dependencies are resolved. It is excellent for a quick replicable test environment, and a super-easy gateway for Mac/Windows users since Docker works cross-platform.
Adriano
@rsvp

Closed #6 ... As a result of pandas revised API for resampling, fecon235 must advance to v5 and require pandas 0.18 or higher.

No more deprecation warnings from using resampling functions: daily(), monthly(), quarterly()