2

I have a Healthcare dataset. I have been told to look at non-parametric approach to solve certain questions related to the dataset. I am little bit confused about non-parametric approach.

Do they mean density plot based approach (such as looking at the histogram)? I know this is a vague question to ask here. However, I don't have access to anybody else whom I can ask and hence I am asking for some input from others in this forum.

Any response/thought would be appreciated.

Thanks and regards.

user62198
  • 1,091
  • 4
  • 16
  • 32

1 Answers1

2

They are not specifically referring to a plot based approach. They are referring to a class of methods that must be employed when the data is not normal enough or not well-powered enough to use regular statistics.

Parametric and nonparametric are two broad classifications of statistical procedures with loose definitions separating them:

  • Parametric tests usually assume that the data are approximately normally distributed.
  • Nonparametric tests do not rely on a normally distributed data assumption.
  • Using parametric statistics on non-normal data could lead to incorrect results.
  • If you are not sure that your data is normal enough or that your sample size is big enough (n < 30), use nonparametric procedures rather than parametric procedures.
  • Nonparametric procedures generally have less power for the same sample size than the corresponding parametric procedure if the data truly are normal.

Take a look at some examples of parametric and analogous nonparametric tests from Tanya Hoskin's Demystifying Summary:

enter image description here

Here are some summary references:

AN6U5
  • 6,808
  • 1
  • 24
  • 42