.NTE0NA.MjYxOTY1

i •v V ■P 0.32.- h0 p 0.1%. £ co 0.2.1»- O-IV- O.lé ■ 0. 12. o.os- 0.01, - ( TL ) ( TL. ^ n f* (n) f (n) eo eo 2 0,001 0,001 3 0,030 0,026 4 0, 150 0,138 5 0,260 0,249 6 0,284 0,283 7 0, 139 0, 147 8 0,081 0,089 9 0,036 0,044 10 0,011 0,013 11 0,004 0,005 12 0,004 0,005 3 if 5 6 7 Ô ^ 10 II 12- 6 : Histograms of number of edges per grain - Expérimental histogram , Histogram correeted by Miles1 procédure. RESUI/P3 MD INTERPRETATION Notice at first that a grain with only two edges is very seldom found. As a conséquence, we have = eQ almost surely. Many methods may be used to interpret the data (multivariate analysis, fitting a model a.s.o.). In what follows, we use the most classical statistical tools (corrélation coefficients, y test) in order to show that things are not so simple as they look. First neighbours can be studied in a satisfactory way. On Figure 6, one can see the histogram f— of the mean number of edges for the first E1 neighbours. ' f— , which has been obtained by averaged measurements, is tighter than ei f . The mean value is E(e.) = 6.48 . eCe.) is larger than 6 since a £o 1 1 grain with many edges is counted many times. The corrélation coefficient between and is - 0.63. It means that the numbers of edges of two adjacent grains are not independent, and more precisely that a grain with many edges often has adjacent grains with only few edges. Regarding the second neighbours, things become quickly inextricable. On Figure 7, one can see the histogram f— of the mean number of edges for the second neighbours.