ESDU 91017:2015
$126.75
Statistical Methods Applicable to Analysis of Aircraft Performance Data
Published By | Publication Date | Number of Pages |
ESDU | 2015-07 | 52 |
ESDU 91017 uses a flowchart to guide its user in the application
of statistical techniques to the analysis of performance data from
which all but random variability has been removed, that is they
have been reduced to standard conditions. After introducing the
concept of hypothesis testing, the possibility of combining small
samples in a larger sample, so improving the accuracy of
predictions, is considered. Fisher's Variance Ratio F test is used
to decide with a certain confidence whether the samples are from
the same population, and Student's t test determines the
significance between sample means. Larger samples, exceeding 30
points, either from combined smaller samples or as measured, are
treated using Normal distribution statistics provided the Chi
Squared Test for normality with a chosen confidence is met. If the
test fails, a transformation of the data (for example, taking the
logarithm) may provide a Normal distribution. The use of
probability graph paper as a means of analysing data to the Normal
distribution is explained. Bartlett's test for the consistency of
variances calculated from three or more samples is introduced.
Regression is considered because, although not a statistical
technique, the results are analysed statistically. The manual
calculation of linear two variable regression is discussed, and
determination of variance about the regression line, the
significance of the regression relationship, regression confidence
intervals, and the correlation coefficient and its significance are
treated. Multiple linear regression is briefly considered. Tables
present the statistics of the Normal distribution, Student's t
distribution, the Chi Squared distribution and the F distribution.
Examples on the use of the methods are found in ESDU 91018 to
91020, 92021 and 92022, 93023 and 96024.