Other Assessment Methods—An Overview
167
this chapter) and the fi tting of a log-geometric fi eld-size distribution
to the observed discoveries. The computational aspects are outlined
in Drew (1990, pp. 147–171). The procedure, described by Drew and
Schuenemeyer (1993), is as follows:
The general form of the parent distribution was isolated by using
a two-stage estimation procedure. In the fi rst stage, the number of
fi elds in each size class larger than the mode are estimated directly
by using a discovery process model. In the second stage, the number of fi elds in size classes smaller than the mode are estimated
by a technique based on inference. Then the two parts are joined
together to construct the total fi eld-size distribution. The technique used to estimate the number of fi elds in the size classes at
and below the mode of the observed distribution depends on recognizing that this part of the underlying distribution is hidden
behind the barrier of cost truncation. This barrier can be removed
by using an inference gained from a study of the collective behavior of the truncation phenomenon across exploration plays and
basins that have different cost/price regimes. Specifi cally, this
inference is based upon the observation that the ratios of the estimated ultimate number of fi elds in successive size classes above
the mode were, on average, constant. The underlying distribution
of oil and gas fi elds estimated by this procedure is log-geometric
in form. The essential component of this distributional form is
that there are more fi elds occurring in each successively smaller
fi eld-size class. (p. 473)
The disadvantages include the following:
A constant ratio between the two adjacent size classes is
1.
diffi cult to validate. For example, the graphs shown by
Schuenemeyer and Drew (1983, Fig. 4) display a random pattern (no trend), but can we conclude the ratios are constant
because they exhibit random patterns?
The statistical assumption, log/geometric distribution, might
2.
not be valid.
Field sizes must be classifi ed.
3.
This method usually presents too large a number of small
4.
pools (Table 7.3).
Examples of this approach are presented in the paper by Schuenemeyer
and Drew (1983).
167
this chapter) and the fi tting of a log-geometric fi eld-size distribution
to the observed discoveries. The computational aspects are outlined
in Drew (1990, pp. 147–171). The procedure, described by Drew and
Schuenemeyer (1993), is as follows:
The general form of the parent distribution was isolated by using
a two-stage estimation procedure. In the fi rst stage, the number of
fi elds in each size class larger than the mode are estimated directly
by using a discovery process model. In the second stage, the number of fi elds in size classes smaller than the mode are estimated
by a technique based on inference. Then the two parts are joined
together to construct the total fi eld-size distribution. The technique used to estimate the number of fi elds in the size classes at
and below the mode of the observed distribution depends on recognizing that this part of the underlying distribution is hidden
behind the barrier of cost truncation. This barrier can be removed
by using an inference gained from a study of the collective behavior of the truncation phenomenon across exploration plays and
basins that have different cost/price regimes. Specifi cally, this
inference is based upon the observation that the ratios of the estimated ultimate number of fi elds in successive size classes above
the mode were, on average, constant. The underlying distribution
of oil and gas fi elds estimated by this procedure is log-geometric
in form. The essential component of this distributional form is
that there are more fi elds occurring in each successively smaller
fi eld-size class. (p. 473)
The disadvantages include the following:
A constant ratio between the two adjacent size classes is
1.
diffi cult to validate. For example, the graphs shown by
Schuenemeyer and Drew (1983, Fig. 4) display a random pattern (no trend), but can we conclude the ratios are constant
because they exhibit random patterns?
The statistical assumption, log/geometric distribution, might
2.
not be valid.
Field sizes must be classifi ed.
3.
This method usually presents too large a number of small
4.
pools (Table 7.3).
Examples of this approach are presented in the paper by Schuenemeyer
and Drew (1983).