est_typing_err: Estimates the Expected Number of Errors in a Data Entry Job

Description Usage Arguments Details Value References See Also Examples

View source: R/est_typing_err.R

Description

Estimates the Expected Number of Errors in a Data Entry Job

Usage

1
est_typing_err(S, F, prior = "postmen", quantity = "mean")

Arguments

S

size of the evaluation sample

F

number of errors in the evaluation sample

prior

can be either 'postmen', 'bad_postmen' or a named vector with prior values. See below for details.

quantity

can be either 'mean', '95q' or 'param'. See below for details.

Details

The function calcualtes (Bayesian posterior) estimates for the expected number of errors (quantity='mean') or the upper bound of the 95% (Bayesian) credible interval (quantity='95q') in the population of entered values when a random sample of S entered values contains F errors. For a discussion on the type of inference, see Winkler et al. (2002).

The calculations are done assuming a Beta-distribution prior over the proportion of errors for all entered values. The users must supply hyper-parameters for the Beta-prior. The function provides two informed default hyper-parameter sets. Both induce an informed prior and are formed from data about postmen's performance in typing post codes as reported in Baddeley/Longman (1978).

The first prior (prior='postmen') says that the a-priori the proportion of errors relative to the total number of entered values is about 1.4% with a variance 0.7. The second prior (prior='bad_postmen') is more conservative saying that it is about 2.1% with a variance of 1.1. Users can also supply their own prior using a named vector that contains the alpha and beta hyper-parameters for the Beta prior (the output of calc_beta_param can be used).

Value

value or vector of the requested quanty/ies.

References

Baddeley, A. D., and D. J. A. Longman. 1978. "The influence of length and frequency of training session on the rate of learning to type." Ergonomics 21(8), 627-635.

Winkler, Robert L. and Smith, James E. and Fryback, Dennis G. 2002. "The Role of Informative Priors in Zero-Numerator Problems: Being Conservative Versus Being Candid". The American Statistician 56(1), 1-4.

See Also

calc_beta_param

Examples

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
 ## Not run: 

 est_typing_err(S=20, F=0, prior='postmen', quantity='95q')

 Example from Winkler et al. (2002) p.3:

 est_typing_err(167, 0, prior=c('alpha'=0.042, 'beta'=27.96), quantity='param' ) 

 
## End(Not run) 

sumtxt/datatools documentation built on Oct. 7, 2018, 11:18 p.m.