This question was written more than 5 years ago. Sometimes we don't look at the past questions, as there are too many present ones. The beauty of this forum is that, instead of bashing you for reviving old threads, we are pleased that you actually went to find the historical questions to find your answer. So here it is, somehow.
The theory behind the calculation of p-value from a confidence interval says:
Get your confidence interval.
Get the upper and lower limits for the confidence interval.
Calculate the standard error as (upper - lower) / (2 * 1.96) (this is for 95% confidence)
Calculate the test statistic as estimate of effect / standard error.
Finally calculate the P value = (1 / √2π ) * e^(-z2/2)
However, getting the confidence interval is quite tricky:
If you are more interested in p-values and related concepts (like confidence intervals) then you may also want to look at the Statistics extension, available in the Marketplace, which provides several other popular statistical tests beyond the t-test and the two-way ANOVA (both of which are available in the basic RapidMiner installation).
Brian T. Lindon Ventures Data Science Consulting from Certified RapidMiner Experts
Answers
This question was written more than 5 years ago. Sometimes we don't look at the past questions, as there are too many present ones. The beauty of this forum is that, instead of bashing you for reviving old threads, we are pleased that you actually went to find the historical questions to find your answer. So here it is, somehow.
The theory behind the calculation of p-value from a confidence interval says:
- Get your confidence interval.
- Get the upper and lower limits for the confidence interval.
- Calculate the standard error as (upper - lower) / (2 * 1.96) (this is for 95% confidence)
- Calculate the test statistic as estimate of effect / standard error.
- Finally calculate the P value = (1 / √2π ) * e^(-z2/2)
However, getting the confidence interval is quite tricky:https://community.rapidminer.com/discussion/44631/how-to-create-confidence-intervals-for-numeric-prediction
Instead of all this burden, you may want to play with T-Test operator to compare two performance vectors using statistical significance tests.
https://docs.rapidminer.com/latest/studio/operators/validation/performance/significance_tests/t_test.html
By using an ANOVA, you'll get the same result. Just take a look at the annotation in the end:
All the best,
Rodrigo.
Lindon Ventures
Data Science Consulting from Certified RapidMiner Experts