Confidence region overlaps the pole. Confidence limits for Amplitude and Acrophase cannot be determined. (CICalc in Cosinor)

Hi, can someone explain to me why this message appears in the CICalc function from the Cosinor analysis?

The message says: " Confidence region overlaps the pole. Confidence limits for Amplitude and Acrophase cannot be determined".

I would like to know:

1. Why the confidence region overlaps the pole?
2. Why the confidence limits cannot be determined?
3. What can I do for this not happen?

Thanks

Answer 1

[Star Strider]

Well, ‘pole’ is new to me, and apparently also to The Dictionary of Circadian Physiology. A PubMed search for it points me to circadian rhythm studies in the high Arctic. Probably not the sort of ‘pole’ you’re interested in.

My guess is that ‘pole’ refers to the centre of the plot. If that’s correct, it would mean the same as the confidence limits including zero, therefore the estimated parameters are not significant, and no circadian rhythm exists for the variable you’re studying.

You might want to email the author of those functions to clarify this.

My only other suggestion is to see if you can fit your data using Curve fitting to a sinusoidal function, but using the Statistics (and Machine Learning) Toolbox function nlinfit and nlparci in addition to (that is, after) the fminsearch call I used. Use the parameter estimates from fminsearch as the initial parameter estimates for nlinfit.

I don’t have your data to experiment with, but it’s worth a go to see if you can get an acceptable fit and parameter estimates. You can then use the cart2pol function to plot it as a characteristic polar plot.

[Kelvin Prado]

Thank you so much Star Strider.

I will try to contact the authors of the function to clarify this issue.

Also thanks for the tip of the sinusoidal function.

[Star Strider]

My pleasure.

The sinusoidal function fit should work if your data are not too noisy. If they are, I would have to tweak it. (I would need a representative sample of your data records, covering at least 24 hours.)

[Kelvin Prado]

In fact, this is not happening in my data. I want to understand this to show a message to the users, so they know what is happening and what they can do to solve it.

[Star Strider]

I admire your altruism, but the function author is responsible for thoroughly documenting their code. If you choose to write an improved version on your own, using your own code, and upload it to the File Exchange, do it!

I never second-guess any File Exchange code, improve it, explain it or troubleshoot it. I figure that is the author’s responsibility. Thus far, I’ve never had to do it with any of my own File Exchange contributions.

[Kelvin Prado]

Thanks for the tips, if I succeed in improving the Cosinor function in any way, I certainly will share it in the File Exchange. I really think we can all work together to write better codes and make our life easier.

[Star Strider]

My pleasure.

My only attempt prompted the routine I quoted. It worked for my data. (I altered it slightly to provide a relevant Answer to that Question.)