Thu, 01/31/2019 - 05:38
Last night I entered 6 visual observations as usual. They showed up at the bottom of the WebObs page as entered, and one of them showed up on the home page as "last observation". (see screenshot). However, I cannot find any of them now by query to WebObs! How could entered observtions "disappear" from the DB?
Are these your "missing" observations?
TT ARI January 30.2188 2018 10.65 LMK
HP AND January 30.2292 2018 16.0 LMK
DA 55 January 30.2444 2018 16.2 LMK
UW PER January 30.2465 2018 16.0 LMK
GK PER January 30.2479 2018 13.0 LMK
HL CMA January 30.2493 2018 12.3 LMK
If so, they are in the database OK. The problem is that you reported the year as "2018" when you probably meant "2019" then when you went to look for them in WebObs Search you likely didn't go back in time far enough to find them. Unfortunately, the "Last observation" listing on the homepage does not give the year!
Believe it not, this is not an unusual occurance near the beginning of each year. Please be careful with this everyone.
Perhaps the observation entry system could warn the submitter of any newly submitted observation more than, say, 6 months in the past.
Ugh, how an innocuous little error in one digit can cause such a "gotcha"! I agree with Eric, that WebObs needs to do some basic checking on the entry fields, especially the time, to warn the user if something seems "way off" !
I find the following procedure helpful. Glance at the JDN for the night for your list of observations at the bottom of the WebObs page.
Then I compare it with the JDN at the top of my entries in my Observing Notebook. As a header for my observations I put in for example "30-31 January 2019 58514". 58514 is the truncated JDN with the first two digits omitted from the integer part.
This specific class of error is already known to be frequent, so this easily automatable check should be automated. People are very often not at their best when submitting data, being either at 3 am or after a long observing night. One's own checks are valuable, yes, as I check my data--but servers don't get sleepy and properly programmed don't miss even one such suspect datum, ever, period.
Automated checks beat "Be Careful" every time. This check should be automated.
Hi Eric & Mike,
I think this would be a valuable check to add to WebObs but first I would like to get some feedback from the community as to what would be a good number of days to use for the check? People do submit older data on purpose so we don't want to stop them altogether or annoy them with numerous pop-ups saying "Are you sure this is correct?".
Would people want such a pop-up to occur when the data are more than 30 days old? 60 days? 180 days? 359 days?
I assume you are including WebObs file uploads as well as individual data submissions, right?
Well, I only submit observations individually, and almost always shortly after doing them. So, my opinion is clearly biased in that respect. For me, any period of time over a few days late should be flagged as "suspicious".
Alternatively, You could add a real-time updated label directly under the date entry field, saying "This observation was made xxx.xxx days ago". If the time is fairly recent, make it in Green, if its several days old, yellow, if it is really old, make it RED :) This won't bother anyone (vs. an annoying pop-up) and it WILL make one take notice if something seems too long in the past!
Mike, I understand your point of view, but I do not belive that a report with observations older than a few days should be considered suspicious just because of the delay.
I can easily imagine several cases when someone may need to report old observations, for example when following the traditional practice of reporting once a month; or in those cases when someone has access to logbooks from an old or deceased observer that never reported his/her observations (of course not to be reported as one's own observations).
I can also cite my own case: I started observing variables in 2015 and went through a self-imposed calibration time period in which I learned to observe using cepheids and Miras as my "school". Most of these observations are not worth reporting, but a few of them were and I did report them when I considered I had been doing well in estimating variables. At that point I did report observations that were over one year old.
I believe that one lesson from your experience (which I also experienced once) is that one has to control one's own reports and correct any of these so called "clerical errors". As others have mentioned, using zapper from time to time to flag discrepant observations is a good way of contributing to the quality of the database.
Cheers and clear skies (we need them badly here in the north ...)
I routinely enter obs a month or two behind. I'd appreciate not having another key stroke entering my 60K + obs