13.09.2014 Views

Published Report (DOT/FAA/CT-94-36)

Published Report (DOT/FAA/CT-94-36)

Published Report (DOT/FAA/CT-94-36)

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

FDADS SIMULATION AT DENVER<br />

A demonstration simulation of the Full Digital Automated Radar<br />

Terminal Display System (FDADS) was conducted November 16 and 17,<br />

1992. Thirty-six 30° blunders were simulated, of which 12 were<br />

shown by later analysis to be at-risk. Three TCV's were found<br />

among the 12 at-risk blunders, resulting in an observed TCV rate<br />

of 1/4. After two days of simulation, the decision was made by<br />

<strong>FAA</strong> management to discontinue the FDADS simulation in favor of<br />

the FMA. The analysis presented here will support that decision.<br />

The analysis will show that the TCV rate for the FDADS simulation<br />

is significantly larger than the TCV rate for the FMA simulation.<br />

The analysis will also show that the larger TCV rate of the FDADS<br />

simulation could lead to an unacceptably large risk.<br />

Although the sample size is small, some conclusions regarding the<br />

data may be drawn. Since the number of at-risk, worst-case<br />

blunders was much larger for the FMA simulation, the observed<br />

Bernoulli ratio for the FMA simulation may be regarded as much<br />

more accurate than that for the FDADS simulation. The observed<br />

ratio for the FMA simulation was 2 TCV's per 186 at-risk<br />

blunders, or 1/93. Since binomial probabilities are easily<br />

computed, the null hypothesis that the probability of a TCV<br />

during the FDADS simulation is the same as that for FMA, Ho:p =<br />

1/93, may be tested against the alternate hypothesis, H1:p ><br />

1/93, directly from the binomial distribution. A significance<br />

level of 0.05 will used to lessen the likelihood of a type I1<br />

error.<br />

If the probability of a TCV during the FDADS simulation was also<br />

p = 1/93, then the probability P of observing 3 TCV's in a sample<br />

of 12 at-risk blunders would be given by:<br />

r2)(lI(gr<br />

= 0.000248.<br />

P = 3 93<br />

The probability of 3 or more TCVIs in a sample of 12 may also be<br />

computed and is found to be P = 0.000254. Since the probability<br />

of 3 or more TCV's in a sample of 12 is smaller than 0.05, the<br />

alternate hypothesis, that the probability of a TCV using FDADS<br />

is larger than the probability of a TCV using FMA, is accepted as<br />

being true.<br />

If the upper confidence limit for the FMA simulation, 5 TCV's per<br />

102 at-risk blunders, is used for the estimate of the probability<br />

of a TCV given an at-risk blunder, then the null hypothesis would<br />

be Ho:p = 5/102 and the alternate hypothesis would be H1:p ><br />

5/102. The probability of 3 TCVIs in a sample of 12 at-risk<br />

blunders, assuming p = 5/102 is computed as follows:

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!