Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

From |
Nick Cox <[email protected]> |

To |
"[email protected]" <[email protected]> |

Subject |
Re: st: Moving Standard Deviation |

Date |
Sun, 3 Nov 2013 16:22:57 +0000 |

I think you are missing my point on divisors. Consider a window with the values 1,2,3,4,5. If you use sample size n (here 5) as divisor the SD is just the root mean square deviation, namely . mata : sqrt(mean(((1,2,3,4,5)' :- 3):^2)) 1.414213562 If you use n - 1 as divisor you get . mata : sqrt(sum(((1,2,3,4,5)' :- 3):^2)/4) 1.58113883 which corresponds to what -summarize- does (I put the same values in a variable) . su y Variable | Obs Mean Std. Dev. Min Max -------------+-------------------------------------------------------- y | 5 3 1.581139 1 5 Ordinarily we can be cavalier about which formula you use but for small windows -- and for reproducibility and comparability with others -- it does make a difference, here a factor or divisor sqrt(5/4), depending on which way you go. -mvsumm- calls -summarize-; so if you want the first definition you need to multiply by a correction factor. Otherwise, sorry, but I have no idea what you mean when you say "is this ok?". Do you distrust -summarize-? -mvsumm-? something else? Whether SD is a good measure of volatility for your project I can't say, naturally. Nick [email protected] On 3 November 2013 14:37, Browne, Alan <[email protected]> wrote: > Hey, Sorry for the late reply. I was under the impression that m was just the number of windows for the moving standard deviation. I am just going to use the mvsumm method as it is far more practical and I was only looking to try and replicate the literature for piece of mind. Just to clarify is the use of mvsumm ok for a moving standard deviation as a measure of volatility? Sorry for the question but trying to figure this out has been driving me crazy over the last few days. Thanks for the help. > > Best regards, > Alan > ________________________________________ > From: [email protected] <[email protected]> on behalf of Nick Cox <[email protected]> > Sent: 31 October 2013 13:28 > To: [email protected] > Subject: Re: st: Moving Standard Deviation > > would be happy with a citation, but my guess is that most reviewers > would regard a citation as unnecessary for something of this kind. > Your notation m perhaps implies that you are using the number of > values as a divisor. If so, you would need to correct the results from > -mvsumm-, which, as documented, uses -summarize-, which, as > documented, uses that number minus 1 as divisor. > > > Nick > [email protected] > > > On 31 October 2013 13:05, Browne, Alan <[email protected]> wrote: >> Hello, I am currently trying to generate the moving standard deviation for a measure of volatility. I have used the mvsumm command, however, I am not sure how to cite that in the paper I am writing. I am trying to reproduce the following formula used in the literature. >> Vol= [1/m*(sum(lne_t+i-1 - lne_t+i-2)^2)]^1/2 >> >> where lne is the log of the exchange rate. Any help with this issue would be hugely appreciated. > > * > * For searches and help try: > * http://www.stata.com/help.cgi?search > * http://www.stata.com/support/faqs/resources/statalist-faq/ > * http://www.ats.ucla.edu/stat/stata/ > > * > * For searches and help try: > * http://www.stata.com/help.cgi?search > * http://www.stata.com/support/faqs/resources/statalist-faq/ > * http://www.ats.ucla.edu/stat/stata/ * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/faqs/resources/statalist-faq/ * http://www.ats.ucla.edu/stat/stata/

**References**:**RE: st: Moving Standard Deviation***From:*"Browne, Alan" <[email protected]>

- Prev by Date:
**st: Conceptual question: treating panel data as time-series or separating the dataset?** - Next by Date:
**st: align the zeros of two different y axis** - Previous by thread:
**RE: st: Moving Standard Deviation** - Next by thread:
**st: Conceptual question: treating panel data as time-series or separating the dataset?** - Index(es):