|
PureBytes Links
Trading Reference Links
|
"Given a problem (question), hypotheses are used to help differentiate
between relevant and irrelevant info with respect to correlation or
causality. The falsifiable requirement does not mean that the
hypothesis is false but that it allows for the opportunity to be shown
to be false..."
Thank-you for the clarification, Bill.:) You've given me a new way of
conceptualizing such problems.
S.
--- In amibroker@xxxxxxxxxxxxxxx, "wavemechanic" <fimdot@xxx> wrote:
>
> Sebastian:
>
>
>
> Given a problem (question), hypotheses are used to help differentiate
between relevant and irrelevant info with respect to correlation or
causality. The falsifiable requirement does not mean that the
hypothesis is false but that it allows for the opportunity to be shown
to be false.
>
>
>
> Your Superbowl stuff is not a hypothesis. You could turn it into one
by, for example, asking whether there is a correlation between who wins
and market direction. Then a possible hypothesis regarding causality
might be that the winning coach is granted power over all specialists on
the NYSE floor and tells them each day whether to take the market up or
down. Is that true or false? After designing an appropriate experiment
you would come up with an answer that would prove or disprove the
hypothesis. On the other hand, a non-falsifiable hypothesis for the
same question might be that an ET on Mars transmits instructions to the
specialists to take the market up or down depending on whether the AFL
or NFL won. To the best of my knowledge, it is not presently possible
to establish whether or not ETs exist on Mars or whether specialists
have the receiving "gear" to get instructions from such an entity. As a
result, this is not a falsifiable hypothesis and is sent to the round
file.
>
>
>
> Similarly, the 200 dma stuff is not a hypothesis but can be turned
into one by asking an appropriate question and then creating a
hypothesis. With respect to why prices rise or fall, one could look at
current theories of price movement based on crowd psychology, formulate
a question about the theory and then a hypothesis that will help answer
the question.
>
>
>
> Bill
>
> ----- Original Message -----
> From: sebastiandanconia
> To: amibroker@xxxxxxxxxxxxxxx
> Sent: Saturday, October 28, 2006 1:53 PM
> Subject: [amibroker] Re: OT: Statistics
>
>
> Bill, I'm not sure what you mean by "falsifiable hypothesis", but
would the Superbowl Indicator qualify?
>
> Some basic background details at:
>
> http://www.snopes.com/business/bank/superbowl.asp
>
> Based purely on the historical "accuracy" rate this indicator is an
excellent one, correct 80% of the time. Clearly there's no causal
relationship between who wins the Superbowl and what the stock market
does, but how could its "success" as a market indicator be disproven
using only the historical data we used in testing?
>
> That's the fundamental problem with testing only using numbers and
not thinking about the "whys."
>
> Now...what's the difference between the Superbowl Indicator and a
200-day moving average? If a stock crosses above a 200-day moving
average it's widely assumed that this means the stock is in an uptrend.
However, a 200-day ma applied to randomly-generated numbers can also
show an "uptrend" when charted. Why is there a causal relatioship with
one but not the other? Who says?:) Using only the historical data
employed by testing, prove it.:)))
>
> So, here we are...there's no way to either prove or disprove if the
200-day ma is a legitimate indicator or not. We know there's no causal
relationship there, the stock isn't going up because the 200-day ma is
going up, it's the other way around.
>
> So what DOES make a stock going up? Whatever that "X" factor is,
why not focus directly on that instead of something that might just be a
meaningless mathematical coincidence?
>
> Just trying to stimulate some thought about what we really know and
what we don't really know. I don't have all the answers, either, I'm
still trying to get rid of all the stuff I "know" that's wrong.:)
>
>
> Luck,
>
> Sebastian
>
Content-Description: "AVG certification"
No virus found in this incoming message.
Checked by AVG Free Edition.
Version: 7.1.409 / Virus Database: 268.13.27/517 - Release Date: 11/3/2006
|