Having just had the first (huge) email of the new-format Mira star 'alerts' one thing stood out (adopts Monty Python 'Spanish Inquisition' voice)... two! Two things:
1) Some folks are observing slowly-varying red stars far too often. Attached is a section of the alert, where all but the red-X's are observations by the same observer - who has observed this star virtually every night in the same week! Miras with short periods such as V369 Cyg or SY Her I personally observe twice a month, otherwise 'reds' really only need one obs per month. PQ Cep has a period of over 400 days, so one obs per month is more than adequate!
2) This isn't a personal thing, but the observer above also added in the 'comments' section the size and focal length of the instrument. This is just unnecessary. I assume 'comments' are notes that might affect the reliability of an observation, nothing else.
Hi Michael,
I have in the past made the attention to these "observers" a few times. Recently I examined the lightcurve of R UMa. I found one observer how made 135 observations of this mirastar in nearly 4 years. But they are not equal over the curve. There are gaps when he is not observing. This way he make an observation every 2-3 days. And almost every estimation is 0.1 magnitude different. This seems to me somewhat cheating. Every time I found an observer like this a I post it at the forum. Regrettably no one from the AAVSO responds to correct the observer. Here is the lightcurve.
Since the thread title is "Overobserving" and not "Miras", I'm going to draw attention to the 373,014 observations since March 2018 on ASASSN-18ey. Almost 1,500/day on one target, and we're fretting over Miras? Seriously?
Recently on this Forum I was overtly berated for going to 2/week for very short-period Miras. The fact that some of these lightcurves have features of a very few days didn't matter to the Holy Church of Once Per Week. What--compared to 1,500/day? Spare me.
So, despite AAVSO's skewed incentives on this: please, let's observe for (anticipated) lightcurve features, not for observation count awards.
As for ASASSN-18ey: The observation request specifically called for time series photometry. And there was a reason for it. The target showed variations of the lightcurve at periods from milliseconds (sic) to days. Seriously.... A crucial part of the lightcurve is not just the fact that the targets flickers, it's actually useful to know how much it flickers and how this changed over the time. And you need multiple observers to distimguisch flickering from random error.
I think it's actually offensive to accuse fellow observers of "cheating" or "overobserving" for whatever non-scientific reasons.
Let me end my comments on ASASSN-18ey with a quote from a recent scientific paper on ASASSN-18ey:
https://arxiv.org/abs/1808.07875 "ASASSN-18ey: The Rise of a New Black-Hole"
"ASASSN-18ey has the potential to be the best-studied BH LMXB out-burst to-date, with more than 40 Astronomer’s Telegrams and >360 000 observations from 50 observers reported on the AAVSO Light Curve Generator as of 2018 October 1"
Doesn't sound like a complaint of overobservation to me.
We are all doing this for fun and for science. There is no such thing as "too much data" in science, never. There are "underobserved" stars, yes, granted. But if an observer, for whatever reasons (training, a private project, whatever) decides to observe a given favorite star once per week, well, that's his or her decision alone how to spend their nights. As long as you are not financing their research, that's none of your business, really, or is it?
HBE
So, whazza problem here? Those "brown marks" seems to line up pretty nicely with the average curve!
Mike
Hi, just taken a look at the detailed LC, and 'your' over-observer is the same as the one I was talking about.
Whether or not his/her "overobservations", are a significant degrading of the data needs to be proven statistically. I think the way this was all going about, starting a long accusatory thread, behind someone's back, is quite the wrong approach. Why not originally send the "abuser" a PM querying abt his methods and reasoning for the frequent obs? Like another poster mentioned, maybe a Mira sometimes does "unusual" things we would all like to know about! Rather than accuse someone publicly, based on questionable evidence. Not the way I would have handled it :(
Mike
The AAVSO is supposed to be a scientific organization, and part of being scientific is realizing that selection bias needs to be discouraged. My own suspicion is that we as an organization overvalue numbers of observations, which has the effect of encouraging us to engage in observing practices that have little scientific value. It seems to me, then, that one possible solution to the over observing problem is to discontinue awards for numbers of observations made. A second solution, probably less controversial, would be to not credit observations by a single observer that have a cadence of greater than, say, 0.05 of a star's period.
I am sure that many of the people who engage in practices such as looking up LPV light curves before each observing session, or observing a year-period Mira once or twice a night think they are making a contribution, when in fact their observations are tainted by selection bias. I am certainly not opposed to looking up the light curve after posting observations, to see how ones' observations compare with those of others, but looking in advance is data snooping and should be avoided.
I am wondering: Why are these topics not addressed in the Observing Manuals? (Excuse me if they are!)
CS
Stephen
Hi,
I have contributed more than 2 million observations to the AAVSO database, hence I am the one with by far (2x at least) the most contributioons.
I also observe some Miras based on a comment that the maxima are no longer visible due to the setting of the star, hence I thought if I cover the minimas one could also use those for period determinations.
So far nobody was interested in the results. Anyway I keep observing for most of the stars I do. I also contributed many 10000 observations to ASASSN-18ey and my data are used by the CBA as the standard to compare other data to.
My data have been used in many publications (search for Hambsch in ARXIV) and I normally observe about 40-50 stars a night in snapshot and time series mode. I cycle several stars in time seiries mode due to the fact that I have a fast mount and fast readout CCD. Sometimes I am even asked to have continuous coverage of a given star (though no MIRA). You can check out the quality of coninuous coverage compare to other observers for e.g. U Ori. I have sometimes MIARs in my field as by product and if I do time series have many observations per night (of course not too useful and needs to be binned to one day averages). Several times my continuous observations have lead to a n improved period (see QX Pup). I see many time that other observers are far off my data and wonder sometimes why this is. However even amongst our Belgiam observers we see differences for the same target due to different filters used and QE of the use CCD camera. I would not consider the light curve of R UMa shown by Hubert as a nice LC due to the scatter at all times and wonder how good the determination of the maximum times is compared to the fact if you would have a decently covered LC like for U Ori. Maybe somebody could do the comparison in using my data and the other data and compare the results.
I am a scientist and that is my two cents.
Josch
Hi Josch an other mega observers, I think you all have missed the point. The discussion is about to many visual observations, not CCD or whatever other photometric observations. That is right, there can't be too many of them. But my point is, estimating the same mira star every 2 days is over observing. You may be sure that such an observer has bias! When such an observer's estimation has every time a decline or increase of 0.1 magnitude, then you know something is wrong.
So remember: this is about visual not photometric.
In 1930-s, an Estonian astronomer Taavet Rootsmäe wrote:
I personally think that every one of us should ask themselves about the motivation to conduct observations of variable stars.
Best regards,
Tõnis
The point that the type of variable matters in this connection is well taken. Even with visual observing, there are some targets where daily observations are valuable. These are not the Miras or other long period variables, but stars like T CrB where things can change suddenly. Yes, the frequent visual observer of T CrB might be biased by his/her earlier observations, but that has to be weighed against missing the start of an eruption. Sometimes I've dithered over whether to visually observe a star like that, but maybe not report all of my observations unless something unusual is happening. But that has a risk of bias, too, so it is probably better to report all observations of possibly cataclysmic stars.
Hi Tõnis,
I am agree with you, what does it mean to the backdrop of eternity, some fast mount and fast readout CCD...
:-)
Velimir
For monitoring purposes, my personal rule-of-thumb is to set my observing cadence to that period of time necessary for a star to vary by my expected error. In that manner, I rarely have two reports in a row that have identical values. Let's take for example a 5-magnitude-range LPV whose cycle is 400 days. That is 50 tenth-magnitude steps. For a visual observer, whose expected uncertainty is on the order of a tenth of a magnitude, you would want to observe no more frequently than 1/50 of the period, or 8 days. This is how the basic observing cadence of a week or two for visual observing of an LPV comes about. IMHO (and not HQ's) it is better use of your time to spread your estimating ability among more stars (there are lots that NEED observations!) than to monitor more frequently than this. For a CCD observer with 0.01mag uncertainties, the observing cadence could be 1/500 of a period, or about once/night. Note that this is on average. A longer period star would have more spacing between observations. Some stars are just nicely placed, and you might want to observe them every night because there is little time penalty, or they are your favorite star. I think there is not much overobserving for the purpose of getting awards these days.
For other types of stars, such as CVs, the variation can be far faster. A CV can rise from minimum to an 8-magnitude maximum in less than a day. For some CVs, you can resolve the 2-hour orbital period. So each star is unique, and the science to be obtained is equally unique. You observe with the cadence requested by the researcher. I did a sub-second light curve of an eclipsing CV with the MMT once, and every point showed a difference from the point before it.
So pay attention to the cadence requested by HQ, and don't be overly concerned if someone else violates that guidance. The AAVSO Target Tool suggests cadences for visual observers, and the observing manuals also suggest cadences, and those are good starting points.
Arne
I was pleased to see HQA's calculation since I sometimes use a similar procedure myself, which often suggests quite different cadences than the various (and sometimes mutually contradictory) online AAVSO guidelines. In fact, I would hope that HQ someday refines its public guidance to acknowledge and incorporate this principle. However - a picayune but important detail - please note that in the example of a 400-day cycle for a 5-magnitude range, the full cycle goes up AND down (or down and up), not just in one direction. This means that the 50 discernible steps are covered twice each cycle (200 days up, followed by 200 days down, on average). So the computed cadence would be once every 4 days, not 8 days. This of course is an approximation since many stars have an asymmetric cycle (faster up than down, or vice versa). In such cases, if the asymmetry is taken into account, the prescribed cadence would be more frequent during the fast-changing part of the cycle, and less frequent otherwise. But, from a purely statistic-theoretic perspective, there is no such thing as "too many data points" as someone already noted, and even multiple observations at the same apparent magnitude have some scientific value.
Amen to that Arne! If you look at the little snippet I attached (and if you got the Miras report) you will notice that all his overobserved stars are well-observed ones as well. He could be observing other stars - and I think thereby getting more personal interest and enjoyment, never mind the science bit - such as some bright DN, which of course DO need daily observing.
I know that in the past I have received certificates for having observed a certain number of stars in a month (I think that's what it was) which maybe is the way AAVSO should go 'acknowledgment-wise' rather than simply numbers of observations - which could encourage 'record-breaking', not that I think the observer concerned was doing that; he may have been acting in perfectly good faith.
Interesting discussion!
I agree with the comments above that it is not desirable nor useful to make many observations so closely spaced in time for slowly varying stars. May be it is just a matter of discussing this topic more frequently, and perhaps reflecting about each one's own motivations and rewards.
In any case I was delighted with Arnes's rule of thumb explained above. It sparked a good deal of thought on how I do observe. Since I started estimating variable stars two years ago I have been using for the stars in my program the cadences recommended here. Arne's message made me rethink this a bit. His criterion seems to me to be more flexible than the usually suggested fixed cadences (which are also good rules of thumb of course!) because it allows some adaptation to the characterisitics of each particular star, and not a class.
The criterion could be summarized as:
dt > (period / dmag) * e
where dt is the cadence (minimum number of days between observations), dmag the given magnitude difference, and e is the error in one's observations. Then, for any given star of known (even if only approximately) period and range, and a given observer's average error one could easily estimate a reasonable cadence that should not be exceeded.
I did some estimates for the stars in my program which includes many red Miras for an average error of 0.15 mag. The results are interesting, for example:
STAR TYPE PERIOD MAX MIN DT
R UMa M 301 6.5 13.7 6
S UMa M 225 7.1 12.7 6
T UMa M 256 6.6 13.5 5
Y UMa SRB 168 7.7 9.2 16
Z UMa SRB 195 6.2 9.4 9
RR UMa M 230 8.6 14.2 6
RS UMa M 258 8.3 14.9 5
RY UMa SRA 310 6.5 7.9 32
R UMi SRB 325 8.5 11.5 16
S UMi M 331 7.5 13.2 8
U UMi M 330 7.1 13.0 8
V UMi SRB 73 7.1 8.7 6
R Vir M 145 6.1 12.1 3
SS Vir SRA 361 6.0 9.6 15
R Vul M 136 7.0 14.3 2
At a glance, the old rule of 7 days for these stars seems pretty accurate too for most stars, but not all. I believe I will experiment with Arne's rule during a while.
Cheers
Hernán [DHEB]
After I wrote this above post I came to think that it would be more correct if the criterion takes into consideration that a Mira star varies by its range in about half its period, which is the same to say that it runs across twice its range in one whole period. So the criterion for a maximum cadence should more correctly be written:
dt > (period / (2 * dmag)) * e
which halves the above given values. This is puzzling, since those values are also about half the suggested cadences. However they make sense, for example R Vir, a rapidly varying Mira, goes from ~11.5 to ~6.5 in about 65 days. This puts the maximum cadence for this star necessary to stay clear a 0.15 mag error in about a day. Of course this is a minimum and can be doubled for good measure, in which case we are again at home in the suggested cadences. The exercise is interesting in any case.
If we are only interested in defining the general light curve period than a strict definition of cadence seems sufficient. However, there are portions of light curves that show within them standstills or humps which may take a different cadence to define them?
Mike
But then the problem becomes one of choosing the best sampling strategy for a dynamic system that does not behave very regularly. That is far more complicated that making a general rule of thumb.
I have read this with great interest and there is another factor to consider. The observing conditions are so variable here in Vermont that it is not possible to use a 'standard cadence' for observing and thus one ends up with gaps in the data collected as wel as runs of data separated by only a day or two.
The point of the discussion is that visual observers need to exercise discipline to avoid the possibility of selection bias. One way of doing that is to apply cadence discipline if observing LPVs. Another is to avoid LPVs altogether - something I do. Here in Virginia, most nights are cloudy, so when a chance comes along to observe, I want to observe as many of my program stars as possible. That is one reason why virtually all my visual observations are of CVs.
On the other hand, my DSLR program includes RR Lyras and Cepheids, some of which I can observe more than once a night. Even with those, I adjust my cadence to not go below about 0.05 period.
Quite so. Yet another thing to consider is that it is unlikely that you are the only observer of a given star in a certain bin period (say 5 days for example) unless - since we are talking about LPVs - your star is for whatever reason very poorly observed. I'm thinking of stars like IN Cyg, ET Aur and so on (this is off the top of my head and I haven't checked that they are in fact poorly observed!) Chances are that someone somewhere has made an obs of 'your' star recently and that will do to add a significant data point to the lightcurve.
So
What's this hangup about cadence here? Certainly that is irrelevant for CCD or DSLR, impossible to human-bias that! For visual observing, there always is a possibility of bias, if one is familiar enough with a given star, that you might get a strong "expectation" of what it should be doing, and may harm your measurement.
Though this is really independent of cadence, rather a matter of poor practice. It's quite valid to make "identical" measurements of a slow changing Mira, every day. As long as you are not biased, and that is the actual best estimate you can make, even though the change may be below your detectability. It's no difference from some other observer making the same measurements as you do, independently, if the same care and precision is used.
I think nobody would complain if 3 separate observers made the same brightness estimate of a Mira 3 days in a row, so why complain if one good observer makes the same identical estimates 3 days apart?
So, the real issue is proper technique. Avoiding "expectation" of where you think a star will be, and do your best to clear your mind, and do every estimate as it was an independent one. That is the real goal every visual observer should strive for, as you gain experience and improve your skills, than worry about "overobserving".
Mike
Perhaps you should familiarize yourself with the work of Twersky & Kahneman. Michael Lewis wrote a very entertaining introduction to their work, called "The Undoing Project". From that, you can find references to some of their actual papers.