I am generating two dimensional graphs using the results of support vector regression analyses. I have 20 independent variables and one dependent. Some of the resulting graphics are good fittings. However, I should have some more knowledge in order to explain these graphs. Does Statistica fix the rest of the independents to some values (for example means) when I get the plot of one independent and the predicted of the dependent variable? If so, this graphs can be taken as sensitivity analyses for the variables.
Thank you for your time ..
Ferda Özdemir Sönmez
Did you generate the 2D plot from the SVM result dialog "Plots" tab?
Below is the example 2D scatterplot generated from SVM regression analysis using Statistica sample dataset Poverty.sta. TAX_RATE is one of the multiple independent variables for SVM regression. PT_POOR is the dependent variable.
X-Y scatterplot generated.
Predicted PT_POOR is computed for each sample case of original dataset using the produced SVM regression model substituted with the observed values of the independent variables for that sample case.
Below is the original dataset with an extra column "PT_POOR Predicted" added. Above 2D scatterplot simply take the "TAX_RATE" and "PT_POOR Predicted" columns and draw a X-Y scatterplot. If you draw plots of other independent variables vs. PT_POOR predicted, e.g. N_EMPLD or PT_PHONE, the same "PT_POOR predicted" column is used.
The method you mentioned like setting mean or special values for other independent variables to calculate predicted dependent variable, does not apply here.
Hi, would you please kindly give us some example graphs that you want to interpret and explain? so that we know what graph you are referring to.
Sorry for the late reply, I was out of city due to semester holiday. I am attaching a sample graph. Thank you for your interest.