Jackknife tool

The previous page was created using a GeoGebra tool called JackknifeResampling(...). This tool can be found in the dropbox folder given in the next page, but here is also a direct link to the ggt-file. The tool takes six arguments:
  • Four numbers/sliders, a, b, c and d. These are typically parameters of the model function. If the model function uses less than 4 parameters, just create 4 numbers/sliders anyway - you don't have to use all of them in the model equation. If the model equation uses more than 4 parameters you would probably be wise to go and revise your underlying theories for your experiment. If you STILL need more than 4 parameters, then study the linear example in the previous section and andjust it accordingly (but manually).
  • A model equation, using 0-4 of the 4 numbers a, b, c and d as parameters. As an example, is a model equation using three parameters.
  • A list of data points with your experimental data.
Using the tool may feel a little awkward. Click the tool, then click on the first number. Then the tool produces popup boxes for the remaining three parameters. Enter the names of the numbers. Then nothing happens since it is waiting for you to click on the model function and the data list. Oncy you have done that, the function produces the following outputs:
  • a list of jackknife functions, i.e. functions fitted to your data set minus one point at a time. If you have 10 data points, this list will contain 10 fitted functions.
  • A matrix of all the jackknife functions' parameter values.
  • The transpose of the first matrix.
  • A list of the averages of the fitted parameters. This list has one value for each used parameter.
  • A list of standard deviations of the fitted parameters. One value for each used parameter.
Now all you have to do to produce confidence intervals is to multiply the standard deviations with the correct factor, 1.96 for 95 % confidence intervals, 3 for 99,7 % confidence intervals etc. It should be noted that using this tool is probably not the best in terms of pedagogy. I think it is better if the students manually remove one point at a time, as shown in the screencast. That will take just a little longer but will help the students understand what they are doing.