[vsnet-chat 234] period error

[Ronald Mennickent <rmennick@phys.cfm.udec.cl>]

Concerning the error estimation: 

a bad thing (unfortunately I did it sometimes)
is fitting a curve to the minimum of the theta window and deriving
the error from the uncertainties of the coefficients. This should reflect
the goodness of the function employed, but not the period's error. 

In general, PDM have not a well defined
statistics, so error cannot be determined in a simple form.
Another best defined algorithm is the AOV (analysis of variance) method
(Schwarzenberg-Czerny, 1989, MNRAS 241, 153). This method provide a
simple manner to determine the period's error from the periodograme profile.

I agree with Dr. Kato that a good alternative is to simulate 
data points adding noise to the original dataset. Applying pdm to every
dataset and analysing their distribution, we could derive,
for example, the mean
period and their standard deviation (error).

A good feeling of the error can be obtained from (order of magnitude)
P*P/2*T, being P the period and T the range of observing time.

Regards
Ronald Mennickent

VSNET Home Page

Return to Daisaku Nogami

---

vsnet-adm@kusastro.kyoto-u.ac.jp