Sourcing Intraday FX Data for R and Managing Data Quality
Summary
The document asks how to obtain accurate intraday foreign-exchange data for analysis in R. One response points to an R package as a way to access FX data directly, potentially avoiding a separate Java download and CSV import workflow. Another response emphasizes that FX trading is decentralized, so there is no single consolidated market record and the price or order flow available varies by participant.
Because data feeds reflect different venues and counterparties, historical observations may not represent a universal market price or the stream available to another trader. The response therefore recommends collecting and retaining one’s own history for the market access being studied. The discussion does not compare package coverage, data costs, sampling frequency, or accuracy, and it gives no benchmark for validating a feed. The package mention is a pointer rather than a detailed setup or data-quality procedure.
Key ideas
- An R package is suggested as a route to FX data without a separate Java-to-CSV workflow.
- Foreign exchange is decentralized, so observed prices and flow vary across market participants.
- Historical data collected from one’s own trading access may best match that participant’s experience.
- The document does not evaluate feed accuracy, coverage, cost, or frequency.
Tags
Full text
# Where and how can I get FX intraday data for use it in R? # Where and how can I get FX intraday data for use it in R? I need FX data (the most accurate possible) for use in R. Right now I have developed a script in Java to download the CSV file (from Oanda) and use this file to read it in R, but I think that is a turn around. Someone know a best option to work with R? ## Answer by user189035 (score 2) https://quant.stackexchange.com/a/30651 Have a look at the TFX R package. ## Answer by madilyn (score 0) https://quant.stackexchange.com/a/29725 FX is a very different beast because there's no centralized market and the actual flow you can interact with differs greatly from participant to participant. I would recommend that you only use historical data that you are storing yourself.
Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.