Jökull - 01.12.1989, Side 68

Fig. 4. The distribution of form parameters. Each histogram (B) and probability plot (A) is based on 1000
grains. The formulas for the parameters are given in the text.
Mynd4. Súlurit (B) og safntíðnirit (A) yfir dreifingu formþátta. Línuritin eru byggð á 1000 kornum hvert.
Lóðréttur kvarði á safntíðniritunum fylgir Gauss líkingu. Formþœttirnir eru skýrðir nánar í textanum.
population with correlation coefficient p = 0.0 are
1.4% and the correlation is significant although very
weak. The level of significance is commonly set at
5% by statisticians, so that the correlation between
elongation and size and between sphericity and size
is significant at that level while the OP-index versus
size correlation is not significant.
FORM PARAMETER STATISTICS
Table II shows the mean value and variance of the
form parameters for each tephra sample as calcu-
lated by the SYSTAT statistics module. The number
of measured grains is equal for all samples (80). The
final line in Table II (Range) lists the critical range
for pairs of means, which is calculated as a part of
the summary statistics by SYSTAT. The value is the
result of the Tukey HSD test at a = 0.05. The critical
range for pairs of means gives the minimum
difference of mean values for any two samples
required to reject the null hypothesis of no
difference. In other words, a null hypothesis stating
that any two samples come from the same popula-
tion of tephra grains can only be rejected at the 95%
level of confidence (a = 0.05) if the difference in
mean values exceeds the critical range. The popula-
tion referred to consists of all tephra grains pro-
duced in all the sampled eruptions.
The results in Table II give some clues to three of
the tests which were included in the objectives of
the sampling programme. The difference in form
parameters for the tephra layer H4 in 4 samples from
one vertical section is smaller than the critical range
and not sufficiently large to reject the hypothesis
that the H4 samples are derived from the same
66 JÖKULL, No. 39, 1989