09.12.2012 Views

I__. - International Military Testing Association

I__. - International Military Testing Association

I__. - International Military Testing Association

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

2E<br />

24<br />

20<br />

I@<br />

12<br />

8<br />

4<br />

Figure 4<br />

Mean Life Comparison<br />

monlna .- --..- - _._ ___ _.____ . _ _. _ __ _ -.-.- .-. -----<br />

0<br />

12 18 20 24 28 32 38 40 44 48 52 68 to 84 T’<br />

months<br />

--- Inalude censors --O 0m1 iill cansors<br />

Figures 3 and 4 show a'comparison between the data base with<br />

all 1000 airmen (event history analysis) and the data base with<br />

700 airmen (i.e., all censored data omitted). The difference in<br />

the two survival functions (figure 3) is greatest at the 48th<br />

month, the point at which censoring is heaviest.<br />

The difference between the two mean life residual functions<br />

(Figure 4) is greatest at the beginning of the 13th month, basically<br />

because excluding the 300 censored data points removes some<br />

information about how long a task is performed. At the 48th<br />

month the two curves become very similar. Thus, censoring after<br />

the first term has less of an effect on the mean life residual<br />

function.<br />

This data could also be presented in a table format. A portion<br />

of these functions is shown in Table 1.<br />

Month<br />

8;<br />

38<br />

39<br />

40<br />

Table 1<br />

Comparison Data<br />

SUrVlval Function Mean Life ReddUal<br />

lndude Omit include cmt<br />

Censora Coneore Censor8 ceneora<br />

A44 -377 13.307 9.273<br />

-627 -364 13.264 8.871<br />

A16 .a40 12.647 8.2U<br />

.496 .a13 12.078 7.969<br />

A79 .293 11.472 7.602<br />

Discussion<br />

The results of this study show that event history analysis<br />

can be used to investigate task perishability. Due to the method<br />

of collecting task data in the Air Force's Occupational Survey<br />

Program, accurate figures can be obtained for the change in state<br />

of the binary variable, e.g., task perishability. Historical<br />

attrition data are available for all career fields. Thus, censoring<br />

is the only unknown variable, and it can be accurately<br />

estimated by combining occupational and attrition data. Therefore,<br />

an appropriate data base can be created for any AFSC.<br />

The results of the analysis also show the advantage of using<br />

event history to analyze task perishability. Figures 3 and 4<br />

vividly illustrate the difference in analyzing task perishability<br />

141<br />

. .

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

Saved successfully!

Ooh no, something went wrong!