Downloaded from https://www.everyspec.com AMCP 706.19 b. Supply voltage The assumption is made in analyzing the data that all measurements are actually from 2. Levels. There are 2 levels of ambient the same population, i.e., factors and levels temperature: have absolutely no effect; this is the null hypothesis. It is foreseen, however, that the a. High mean lives under certain conditions may not be the same; this is then taken into account b. Usual in calculating the various sample variances. Then we see, according to the actual data, and 3 levels of supply voltage how likely we were to get the results we got. If the results would be very unlikely-say a.
High less than I chance out of 1000-we usually then reject the original assumption. The b. Usual details of the analysis are more complicated, of course. c. Low As the name implies, we estimate the 3. Replication. There are 4 receivers variance of the data in several ways, and operated at each possible condition. then compare the variance estimates. In order to make the comparison feasible, the Table 2-14 shows the measurements which variances must be estimated in a particular will be made. This is called a full-factorial way. All of the estimation and analysis can experiment since all possible combinations of be performed without the assumption of s- levels and factors occur.
Full-factorial ex- normality, up to the point of making quan- periments are often too expensive and time titative s-significance statements, consuming to run. Suppose that: Suppose a population has a variance ov I. The average life of a receiver under the experimental conditions is 3 months. TABLE 2-14 2. There are 6 test stands, each costing $10,000. EXPERIMENT ON RADIO RECEIVERS Each x represnts a rnewurement of lif.. 3. Time on a test stand costs $1000 per month. supply Voltpaj Low U"ua Hiluh It is readily seen that the experiment will consume a calendar year and will cost Usual x x x $132,000 just to set it up and run it--as- x x x sumbig nothing goes wrong.
Ambient x x x Temperature I x . x Fractional factorial experiments are dis- High x x x cussed in texts on experimental design. This X x X is a sophisticated subject and requires a X X x knowledge of statistics and engineering. X X x Only full factorial designs are considered in h w. 24 (2 X 3 X 4 - 24) menwemrents to be the remainder of this paragraph. me, 2-79