Invalid Results Gallery

Introduction

Due to poor conditions - such as high levels of RFI - sometimes the results of the automated observation and analysis process are invalid.  Inconveniently these invalid results can sometimes look like a 'glitch' in the 'Measured Period Offset' plot as shown in the example below - where a 'glitch' of ~ 1.5 ppm appeared to have occurred. The subsequent observational results after this result, i.e. the two blue dot data points to the right of the invalid result where the measured period offset returns to 'normal', show this is clearly not the case.

Identifying Invalid Results

Fortunately in most cases the invalid result can be identified immediately (i.e. without needing to wait for the next day's observation) by examining the accompanying PDMP plot. The example observation which produced the invalid 'glitch' result is shown below - where it is clear that either there has been a high-level of RFI or - possibly - an extreme drop in flux density due to refractive scintillation (unlikely as Vela is not known to exhibit this) - making this a clear invalid result.

Sometimes - rarely - the 'bullseye' is reasonable and the reported S/N is passably high and yet it is not certain that the 'glitch' it reports is valid. A following observation is useful for ruling out a 'glitch' event. An example of that scenario is shown below.

The analysis application PDMP has returned a passable 'bullseye' and reasonable 'S/N' (11.37) but has obviously 'lost the plot by returning a 'glitch' of magnitude about 2 ppm as shown below.

The key to this false positive is the heavy RFI near the end of the observation (in the 'Phase vs Time' plot).

Also - near the 0.2 mark in the 'Phase vs Time' - there is a faint vertical trace which most likely is the real pulse - although strangely there is nothing discernible in the 'Phase vs Frequency' plot at the same phase. That's a bit of a puzzle.

The following day's observation result showed a return to the expected spin-down period and cements the previous result as invalid - so - no 'glitch'. Careful examination of the result below should reveal to the reader the new point near the expected period from the following day.

Invalid Result Action

When an invalid result is identified it is subsequently expunged from the data record so as to not contaminate the statistical analysis. This is usually done within a few days of the date of the invalid result.

An example of such action is shown below (08 May 2023) - where the invalid result as shown above (07 May 2023) has been removed from the dataset.

NOTE: as can be seen from the above graph, there remains several other 'outliers' where the measurement has returned an obviously erroneous result. It is tempting to remove those - but as there is no apparent reason for this deviation in those cases - other than random noise - they should not be removed. The removal of data for the reason of 'random noise' runs counter to the scientific method.