PureBytes Links
Trading Reference Links
|
i am exploring some typical price-time series data
problems ( departure from normality features )
and how those may affect typical
normality when modeling financial time series.
if somebody has knowledge of issues of:
- price discreteness
- liquidity / illiquidity
- sparse data
- bad ticks
and how those affect typical price-time data normality
assumptions, i would appreciate if you could either
point to papers on those issues or e-mail back with
your experience.
for example,
we can observe the following data features on 1 min 30 year US bond futures:
- discreteness - yes, data is highly discrete
- liquidity - it is pretty high
- sparse data - front month has little sparse data, only towards rollover or on low
volume days.
- bad ticks ( outside trades ) - almost none, however many pop and drop
long range bars resemble bad ticks ( outlier bars )
or we can observe the following data features on 1 min DIA:
- discreteness - yes, to a degree
- liquidity - low
- sparse data - yes, a lot of it
- bad ticks ( outside trades ) - frequent
if we however look at the daily or 30 min data of those two symbols
we can observe that those features are pretty much ALL gone.
if modeling risk for instance if those features are not accounted for
( on high resolution data for example ) the model may fall apart
if data is not within the "normal" bounds... the trading system will bomb,
unless specifically adjusted for those features.
reason i am interested is that those is
i have a pretty good understanding of those but need to check
myself against the consensus to make sure i am not missing anything.
so, experience, papers, books??? anyone?
bilo.
ps. i would also be interested in real time implementation
of bad tick ( quote ) checking vs automated trade execution.
|