Value Based Triangle Testing

• Can be too “statistical” for business partners
• Alpha, beta, % confidence, power, p-value, d-prime, etc…

library(sensR)
discrim(37,100,method="triangle")

Estimates for the triangle discrimination protocol with 37 correct
answers in 100 trials. One-sided p-value and 95 % two-sided confidence
intervals are based on the 'exact' binomial test. 

        Estimate Std. Error Lower Upper
pc         0.370     0.0483 0.333 0.472
pd         0.055     0.0724 0.000 0.209
d-prime    0.643     0.4378 0.000 1.318

Result of difference test:
'exact' binomial test:  p-value = 0.2489 
Alternative hypothesis: d-prime is greater than 0 

• May seem contradictory to business: same % correct (41.7%), different conclusions
• We are easily enslaved by p-values (i.e. “significant”“ if p < 0.05)…

Estimates for the triangle discrimination protocol with 20 correct
answers in 48 trials. One-sided p-value and 95 % two-sided confidence
intervals are based on the 'exact' binomial test. 

        Estimate Std. Error Lower Upper
pc         0.417     0.0712 0.333 0.568
pd         0.125     0.1067 0.000 0.352
d-prime    0.991     0.4590 0.000 1.813

Result of difference test:
'exact' binomial test:  p-value = 0.1423 
Alternative hypothesis: d-prime is greater than 0 

Estimates for the triangle discrimination protocol with 60 correct
answers in 144 trials. One-sided p-value and 95 % two-sided confidence
intervals are based on the 'exact' binomial test. 

        Estimate Std. Error   Lower Upper
pc         0.417     0.0411 0.33515 0.502
pd         0.125     0.0616 0.00273 0.253
d-prime    0.991     0.2650 0.14087 1.475

Result of difference test:
'exact' binomial test:  p-value = 0.02237 
Alternative hypothesis: d-prime is greater than 0