box and percentile question

Hi everybody, does anyone know if the percentile panel included in Igor take NaNs in waves into consideration when calculating the percentiles? Or the procedure remove the NaNs by itself? i.e. I don't know if waves should be cleaned to include no NaNs for this panel.

Thank you!
Well, I wrote that but couldn't remember... but a look at the code indicates that it looks for NaNs and compensates for them. Should be easy enough to test by experiment...

John Weeks
WaveMetrics, Inc.
support@wavemetrics.com
Thank you johnweeks! I did a test a one wave and manually created some NaNs in the wave, the panel gave me back different results so I suppose there is a difference in here? I have one more question though, how should I find the procedures for this panel? Or what would be the correct way to locate procedures that are integrated in the software and the menus? Plus, what does the suffix _EB stand for in the output from the percentile procedure? Thank you very much!
Are you comparing a set of waves without NaNs to a set of waves where some data points have been replaced with NaNs? If so, then the result must be different- the NaN points are simply left out; that is, if you have 5 waves and a NaN in a row of one of the waves, then the percentiles for the row with a NaN in it will be based on 4 values, not 5.

John Weeks
WaveMetrics, Inc.
support@wavemetrics.com
mwpro wrote:
I have one more question though, how should I find the procedures for this panel? Or what would be the correct way to locate procedures that are integrated in the software and the menus?

Windows->Procedure Windows will give you a menu of the procedures. Some WaveMetrics procedures are implemented a Independent Modules, which are hidden by default, but Percentile and Box Plot.ipf is not one of those.
Quote:
Plus, what does the suffix _EB stand for in the output from the percentile procedure?

When you make a box plot, the boxes are implemented as error bars. Those are the waves that define the error bar boxes.

John Weeks
WaveMetrics, Inc.
support@wavemetrics.com
Thank you johnweeks! Creating NaNs does make it a different wave, so they will have different percentiles, and now I understand that I don't need to remove any NaNs from the target wave since the inherent procedure does it for you. Thanks again!