As one would suspect the more evenly spaced gauges placed on the surface of a sample, the more accurately the average gauge value will predict the true surface strain mean, as can be seen in Figures \ref{481059} to \ref{890019}. Comparing the relationships of the splitting test and 4 gauges with the real surface mean, (Figures \ref{752700} and \ref{404259} respectively) the results are quite comparable. The splitting test produces a mean standard deviation of \(262\ \mu\epsilon\) and the four strain gauges of \(260\ \mu\epsilon\) in the full populations set and \(132\ \mu\epsilon\) and \(128\ \mu\epsilon\) in the limited set, while using eight gauges produces an advantage over both \(\left(184,\ 100\ \mu\epsilon\right)\).
The results for the full and limited populations are presented in Table \ref{542582} with maximum, minimum and mean values. Of particular interest is the limited population set perpendicular splitting test standard deviation (the standard deviation of the difference between the two splitting tests performed on the same sample) is \(244\ \mu\epsilon\), which as a 95% confidence interval is \(478\ \mu\epsilon\), while this is lower than the \(589\ \mu\epsilon\) found using the experimental method in Chapter 5, it may just be an indication that some of the populations with fairly consistent surface strains are over represented compared to the Chapter 5 data (or that the experimental data was incorrectly specifying some measurement error with rotational error). Biologically it seams unlikely that completely consistent surface strain profiles exist but a lower bound on 95% confidence intervals of approximately \(\pm480\ \mu\epsilon\) on the repeatability of a rapid splitting test on samples similar to those represented here seams reasonable (ignoring measurement error). The lower bound (it is likely to be less acurate) on the 95% confidence interval of the prediction of the real surface strain mean from the rapid splitting test is \(\pm\ 281\ \mu\epsilon\). Given the limited population set requires a mean input surface strain of \(877.7\ \mu\epsilon\) to get an output population with a standard deviation of \(630\ \mu\epsilon\) the intra and inter stress variances are partitioned approximately in half i.e. the standard devation within a stem \(\left(611\ \mu\epsilon\right)\) is approximately equal to the standard devation between stems \(\left(630\ \mu\epsilon\right)\), within the limited population set. As above however, the within stem variation is probably a little higher in the experimental work from Chapter 5.