topic Re: Get local fits after Kernel Smoothing in Discussions

Get local fits after Kernel Smoothing

Teresa_ — Tue, 26 Nov 2019 08:34:23 GMT

Hi all,

I smoothed my data with the Kernel Smoother / Local Smoother (Local Fit: Linear, Weight funtion: tricube). My aim is now to extract the equations of all the local fits, which was done by the Kernel Smoother.

Is there somebody who could give me some advise on how to best set this up?

I am really struggeling with this problem. I would be really grateful if someone could help me here.

Thank you in advance for your help.

Re: Get local fits after Kernel Smoothing

wgardnerQS — Tue, 26 Nov 2019 14:07:53 GMT

If you are using the smoother in graph builder:

Red triangle > show control panel
Under the smoother section that appears below the column list
Red triangle > save formula

If you are using the smoother in fit y by x:

Red triangle (next to "Local smoother" under the chart) > save prediction formula

Either of the above will add a formula column to the datatable as the rightmost column, containing the smoother formula

-Will

Re: Get local fits after Kernel Smoothing

martindemel — Tue, 26 Nov 2019 14:19:37 GMT

Using Big Class and Bivariate Fit of height by weight. You can use the kernel smoother and get a fit report as well a red hotspot next to the fit legend:

There you can save the coefficients (X Loess, Y Loess), as well as the prediction formula. The formula will be harder understand if you just look into the formula editor

, but double clicking in the formula will show you the full formula:

Interpolate(
	:height,
	[51, 55, 58, 60, 62, 64, 66, 68, 70],
	[66.4237610962498, 77.9887711985387, 87.676320121761, 99.4642495608787, 100.773120401001, 105.139185406223, 117.30361698686, 130.514041992402, 144.259907247776]
)

The values used are the coefficients again. Based on your sampling Delta you may get more/less coefficients.

Is that what you are looking for?

Re: Get local fits after Kernel Smoothing

Teresa_ — Tue, 26 Nov 2019 14:45:45 GMT

Hi Martin,

I think this is not exactly what I am looking for.

With the linear Kernel Smoother every area is fitted with the equation y=a+bx. And I want to get the values at least for b.

Is it possible to get these values?

Thanks a lot for your help!

BR, Teresa