1-10. CALCULATOR EVALUATION USING STANDARD DEVIATION

TM 5-6635-386-12&P the counts will equal the square root of the average count or mean. Given One standard deviation equals 68% of the population of events for a series. Put these all together and: + 1 Std Dev = + - = Averagecount 68%Prob. This finding can be put to good use in evaluating tester performance if we use it to inspect a series of counts taken in the standard count mode. (The procedure may also be used to evaluate any series of nuclear tester counts. It is only necessary that the tester not be moved during the series of counts. Counts could be taken in any operating position, with another tester nearby, etc.). Two evaluation methods will be described. One uses a hand calculator with statistic functions, the other uses a manual checkoff method. Both techniques work well. Both should be used with ten successive counts taken in the 1/4 minute time period. (This will not work if counts are taken in other time periods due to the averaging or normalizing of the tester on long time periods). 1-10. CALCULATOR EVALUATION USING STANDARD DEVIATION Enter the series of numbers into the calculator following calculator instructions for standard deviation. Determine the deviation and place it in one of the calculator’s memories. Pull up the mean of the series of counts and determine the square root of the mean. Divide the deviation by the square root. From the above discussion, if the series was "perfect", the square root would equal the deviation and the division would be "1.0". The series will seldom work out perfectly, however, and some variance from 1.0 will be observed. It is normal for this to lie between 0.75 and 1.25 with a general tendency towards 1.0 for a normal tester. 1-11. MANUAL EVALUATION Take the same series of ten numbers. Use 1/4 minute time key. If the standard deviation of the series is to lie within + - average and this is to include 68% of the series, then this is merely stating that 68% of the numbers should lie within + - of the average or mean and that 32% should be outside the average or mean. That is, out of a series of ten numbers, 32% or about 3 out of 10 will lie outside plus or minus the square root of the average. We simply add up a series of numbers, average them, determine the square root of the average and then add and subtract this square root to the average. The resultant high and low limits will include 68% of the numbers in the series and 32% will be higher or lower than the limits. The following example illustrates a typical tester placed in Standard Count configuration and using the Student Field Data Worksheet (DA Form 5448R (Moisture and Density Tester Field Data Worksheet)), Figure 1-3, to accumulate the data for ten counts. DA Form 5448-R is located at the back of this TM for local reproduction authority. This tester was normal in all respects. We seldom are fortunate enough to have the series work out exactly 3 out of 10 each time we run a series. However, if we make an intelligent allowance for variations in numbers we will observe that the series will exhibit a general trend towards 3 out of 10. An occasional 2 out of 10 will be observed and an occasional 4 out of 10 will be observed. However, the statistic probability of a 5 out of 10 or a 1 out of 10 is very slim, and such splits should be very rare for a normal tester. It is important to observe the average from one such Standard Count Evaluation to another. If the average is within approximately 1/3 to 112 of the square root of the prior average, then the difference between the two series is normal. 1-10