• No results found

a) Introduction

Bivariant statistics describe the relationships between

pairs of variables. A variable in this case is either the amount

of a major or minor element in a sample. In most cases the most

useful device for the presentation of the data is a scatter

diagram. Usually, the data can be quantified in terms of

regression and correlation coefficients.

The coefficient of correlation between two variables that

is conventionally used in scientific studies is the product moment

correlation coefficient. There are however a number of objections

to using this coefficient. These are discussed in Appendix C. In

this work the Kendall rank order correlation coefficient is also

used because it is less sensitive to the constraints that apply to

the former.

b) The correlation matrices

The correlation coefficients for both methods of

calculation are presented in Table 3-10. This is a matrix for all

combinations of variables. The matrix is based on the maximum

possible number of comparisons within the data set eg. the

coefficient for Si with Al is based on 107 points and that for Si

with C on 60 points. It should be noted that the values of the

coefficients may be affected by the phenomenon of closure which

arises when data are referred to a constant total such as 100 in

55 -

coefficient calculated from such data may be quite anomalous, with

the effect being most marked between the most abundant variables

(Chayes 1971). Caution must accordingly be exercised in the

interpretation of such matrices.

It is pertinent to observe here that when the two

coefficients of correlation are compared, the sign is the same for

each but in most cases the degree of estimated correlation of the

rank order correlation coefficient is lower than that of the

product moment coefficient. There are exceptions however, eg. the

correlation coefficient calculated between K and Al is 0.48 when

calculated by the product moment method and 0.60 when calculated

by the Kendall rank order method.

Sulphur has not been included in the correlation matrix

because only 27 analyses are available. It however correlates

well with Fe. Correlation analysis provides a product moment

correlation coefficient of 0.81 and a Kendall rank order

correlation coefficient of 0.50.

c) Element associations

Most of the comparisons yield positive correlations except Si,

which has negative correlation coefficients with all other elements

except C. The strongest correlations generally appear to be

associated with the major elements; this may be partly due to the

closure phenomenum. If an arbitrary figure of 0.50 is chosen for

the Kendall rank order correlation coefficient, it is found that

the linear regression for all element pairs exceeding this value

and regression lines for those element pairs which have a

correlation coefficient equal to or greater than 0.50 are presented

in Figs. 77-85.

There are 31 strongly correlated pairs of elements out of

a possible 136 pairs. Although Sr is not well correlated with Mn,

Mg and Ca when the total sample suite is considered, it is found

that when only those samples from Dobb’s Linn are considered,

these elements correlate well. Correlation analysis for the

Dobb’s Linn sub-set of samples yields Kendall rank order correlation

coefficients of 0.55, 0.53 and 0.59 for the variation of Sr with

Mn, Mg and Ca respectively. When these correlations and that of

Fe with S are included, there is a grand total of 35 correlatable

pairs of elements within the data set (Fig. 86a).

It is possible to assign most of the better correlated

element pairs to one of four groups (Fig. 86b), The first group

consists of all ten possible combinations of the elements Al, K\

Ti, Rb and Zr. Attention has already been drawn above (Section 6b)

to the similarity of the vertical profiles of the elements in this

group. The second group consists of all six possible combinations

of the elements Ca, Mg, Mn and Sr. This group of elements is

most significant at Dobb’s Linn. The third group of elements may

be divided into two parts. The first part consists of all six

possible combinations of the elements Fe, Al, Mg and Zn whereas

the second part consists of two of the three possible combinations

between the elements Fe, S and Cu. The fourth group is composed

of negative correlations between Si and the elements Ti, Al, Fe,

- 57

to any of the four groups are Zn with Mn and Zr with Mg.

The first three groups, with their high positive internal

correlations, are clearly very significant and are considered

below in greater detail. The fourth group with its negative

correlations based on Si, may be attributed to a greater or lesser

extent, to closure. Where elements are presented in a given

mineral species in approximately constant proportions, high

positive correlations may reflect the variation in abundance of

that mineral in a range of rocks.

The regressions for all 10 combinations of the Al - K -

Ti - Rb - Sr group pass close to the origin and suggests that the

majority, but not all, of these elements are found in one mineral

species. X-ray diffraction studies have confirmed the presence

of a mica conforming to the 2M type of the illite group. If the

element associations are indicating this mineral species, then the

mica must contain trace quantities of Rb and Sr.

The approximate proportions of each element within the

mineral species have been calculated from the slope of the

regression lines. A composition of Al 75.6%, K 19.1%, Ti 5.0%,

Rb 0.1% and Zr 0.2% resulted. These proportions are not

dissimilar to those of the type Fithian illite which contains

71.5% Al, 26.2% K and 3.7% Ti (Weaver and Pollard 1973). This

lends support to the diagnosis of illite.

In the sedimentary environment the elements Ca, Mg, Sr

and Mn are commonly associated as components of carbonate minerals

X-ray diffraction studies have shown that only a few samples have

that Mn correlates to a limited extent with Fe (Fig. 83). The

regression slopes for the correlation of these five elements

indicate the presence within the Moffat Shales of a carbonate with

63.1% Ca, 7.2% Mn, 25.7% Mg, 1.4% Fe and 0.1% Sr.

Element proportions calculated from the regression slopes

for the third group of elements are not as meaningful as those

computed for the first two groups. This is partly due to the link

through Al with the group one elements, partly to the common link

through Fe with the group two elements and partly to the common

link with Fe within the group (Fig, 86a).

The regression slopes of the Fe - S - Cu sub-group indicate

the presence of a pyritic mineral with 42.3% Fe, 57.2% S and

0.4% Cu. Similarly an iron-rich chlorite, related to clinochlore,

of composition Fe 36.3 : Mg 14.7 : Al 49.0 : Zn 0.05, is indicated

by the regression slopes of the Fe - Mg - Al - Zn sub-group.

d) Association of elements with minerals

The variation of element concentration has been correlated

with mineral content as calculated from X-ray modal analysis.

Product moment and Kendall rank order correlation coefficients are

given in Table 3-10. The correlation matrix indicates 13 mineral-

element pairs with a Kendall rank order coefficient greater than

0.50. Not included in the matrix is the variation of sulphur with

pyrite. This combination gives a product moment correlation

coefficient of 0.81 and a Kendall rank order correlation coefficient

of 0.50.

59 -

illustrated in Fig. 86c and scatter plots with superimposed

regression lines are given in Figs. 87-89. It is observed that

Si exhibits negative correlations with albite and sericite but is

positively correlated with the amount of quartz in the sediments.

Al, Fe and Mg yield negative correlations with quartzjand Ti, Al

and Mg are positively correlated with the amount of sericite.

Albite, although not recognised as a distinct mineral species when

considering element associations, is found to correlate with both

Al and Mg. Pyrite, iron and sulphur intercorrelate lending

confidence to the observed vertical mineral distribution

(cf Figs. 22,49,57). Correlations within the mineral group itself

indicate that as the amounts of sericite and albite increase

within the sediments there is a decrease in the amount of quartz

present.

In certain horizons within the sequence, notably within

the M.cyphus Zone, pyrite is visible in hand specimen. This

observation plus the good statistical correlations between Fe, S

and modal pyrite indicate that Fe is principally bound up in the

sulphide phase. Cu, Pb and Zn can be present in pyrite in the

form of admixtures of other minerals such as chalcopyrite, galena

and sphalerite (Fleischer 1955) or as small amounts in solid

solution with pyrite (Deer et al,, 1962).