Missing data can arise in the raw data or as a result of processing of the data. In the former, which may be for example because of telemetry problems, the pixels are given a value of -1. Once any processing is done on the dataset, missing or corrupted pixels are given the value -100. The cosmic ray cleaning procedure cds_new_spike, for example, by default fills in bad pixel values with a value derived from surrounding pixels. For the purest who may object to 'inventing' data it also has the option to label affected pixels with the missing data value.

Many of the display routines need to know what value missing data are set to so they can safely ignore those pixels in their processing. Most of these routines (described in CDS Software Note #5 PostScript - 78 Kb) access a system variable to determine this value. This system variable can be set by the command:

• IDL> setflag, missing=value

To see the effect of this try:

• IDL> arr = dist(100,100)
• IDL> plot_image, arr
• IDL> arr(indgen(50)*10) = -1000
• IDL> plot_image, arr
• IDL> setflag, missing=-1000
• IDL> plot_image, arr

Some general analysis routines cannot use the system variable and must be explicitly told what data to treat as missing. For example the average routine.

• IDL> print,average(arr,missing=-1000)
• IDL> print,average(arr)

would give very different results.