Is this usefull for you?

http://mnras.oxfordjournals.org/content/377/2/677.full

I do not see that this paper has anything to do with the question.

I am observing in a team intensively HADS but only in V band. Are your V and R band data of the same quality? A worse uncertainty could maybe fool the period search algorithm for the different bands and affect the smaller amplitudes more than the prominent one.

Just a guess. More information would be needed.

Regards,

Josch

I have never used Period04, but a glance at its User’s Guide indicates that it can accept input weights for each magnitude sample. If you can supply realistic weights (maybe you did, but it is not mentioned), then apparently Period04 can give you what it calls “sigma” values for frequency, amplitude, and phase. For example, if your measurement errors can be approximated as Gaussian, then the weight for a given magnitude measurement would be the inverse of the squared uncertainty sigma. If you have reliable photometric zero points, it might be better to work in flux units instead of magnitude units, since Gaussian random variables may be negative, while negative fluxes don’t work well with magnitude conventions, and clipping flux to some positive value biases the noise. This can be significant for faint sources but can usually be ignored when the flux is large compared to the uncertainty.

I would be interested first in the ratio of amplitude to amplitude sigma for all the extracted frequencies, i.e., a S/N ratio. It may be that the discordant frequencies are not actually statistically significant. Whether they are, and if so by how much, is extremely important to establish.

This is an example of why uncertainty analysis is so important in science. The error estimate for each measurement must be realistic. If you overestimate your errors, real variability will appear not statistically significant, whereas if you underestimate your errors, simple noise fluctuations will appear significant. In any case, I recommend being as careful as possible about the role of measurement errors in your data.

If you really have significant differences in variability frequencies, then there are probably some interesting physical processes going on. It may be only two separate unresolved sources, or if truly a single source, perhaps the R band is showing some time-dependent activity in a surrounding dust cloud. The results could be of considerable interest if truly significant.

Thank you very much.

I agree, this is an exciting result if, as John Wendell Fowler mentioned, it is statistically signficant.  It would suggest that the frequencies in the two different bands are reacting to different physical variables.   Might it be worthwhile to compare the pulsations in these differing frequencies with those made of similar stars about which more is known of their environment?  Just a thought -- I'm hoping you will post a citation to your findings here.