power_z_cor() can return negative probabilities

[LukasRoeseler]

When using extremely small values, negative values for power are reported that violate the probability axioms by Kolmogorov. Should they actually stop at alpha or at zero?

Reproducible example: TOSTER::power_z_cor(n = 10, rho = 0, power = NULL, null = .001, alpha = .05, alternative = "equivalence")

Output:

> TOSTER::power_z_cor(n = 10, rho = 0, power = NULL, null = .001, alpha = .05, alternative = "equivalence")

     Approximate Power for Pearson Product-Moment Correlation (z-test) 

              n = 10
            rho = 0
          alpha = 0.05
           beta = 1.799998
          power = -0.7999976
           null = 0.001, -0.001
    alternative = equivalence

[arcaldwell49]

Good catch. I'm actually surprised the function doesn't fail in this case given how it works...

At the very least, the answer should be "0" for a TOST procedure like this.