Estimating Mature Plays
53
The standard deviation is typically larger than the mean, which
3.
makes it less useful for prediction and comparison.
The interval from the 75th (8.6 × 10
6 m
3
) to the 25th upper percentile
(16.4 × 10
6 m
3
) is a 50% prediction interval for the pool that contains the
median. That is, the probability that the pool will have a value between
8.6 × 10
6 m
3 and 16.4 × 10
6 m
3 is 0.5. Similarly, 5.5 × 10
6 m
3 to 26.4 × 10
6 m
3
is a 90% prediction interval for the largest pool. The latter prediction
interval has a higher probability of occurrence, but at the expense of
having a much wider interval (i.e., more uncertainty).
In the following discussion, we will start with the 75% to 25% prediction interval as a statistical measure of goodness-of-fi t, and the median
will be used as a point estimator of pool-size-by-rank. The 75% to 25%
interval was derived from pilot studies. In cases when the 75% to 25%
interval does not match most or all of the discoveries because of the
presence of large outliers, the 95% to 5% interval should be used to
match current ones.
Matching proceeds as follows: Point estimates derived by LDSCV
or NDSCV are used fi rst to predict individual pool sizes because they
generally yield reasonable estimates. In cases when the point estimates
do not predict all discoveries because variances are underestimated
and/or the means are overestimated by LDSCV or NDSCV, the estimated values of μ and σ
2 are chosen from their prediction intervals and
reapplied in the matching process to predict all discovered pool sizes.
The procedure for fi nding μ and σ
2 values is as follows:
Examine which
1.
μ and σ
2 can accommodate most discoveries
with the 75th to 25th or 95th to 5th percentile interval.
Determine the degree of best prediction by measuring the
2.
minimum distance between the discovered pool size and the
median or mean of the predicted pool size.
Verify statistical predictions by examining their geological
3.
implications. After each individual pool-size-by-rank prediction, we can observe whether the implications conform to the
geological model.
Examples of the types of question that one should ask after each prediction are: Have we discovered the largest pool yet? What are the sizes
of the remaining largest pools? What is the potential of the remaining undiscovered pools? Have we predicted enough small pools for the
53
The standard deviation is typically larger than the mean, which
3.
makes it less useful for prediction and comparison.
The interval from the 75th (8.6 × 10
6 m
3
) to the 25th upper percentile
(16.4 × 10
6 m
3
) is a 50% prediction interval for the pool that contains the
median. That is, the probability that the pool will have a value between
8.6 × 10
6 m
3 and 16.4 × 10
6 m
3 is 0.5. Similarly, 5.5 × 10
6 m
3 to 26.4 × 10
6 m
3
is a 90% prediction interval for the largest pool. The latter prediction
interval has a higher probability of occurrence, but at the expense of
having a much wider interval (i.e., more uncertainty).
In the following discussion, we will start with the 75% to 25% prediction interval as a statistical measure of goodness-of-fi t, and the median
will be used as a point estimator of pool-size-by-rank. The 75% to 25%
interval was derived from pilot studies. In cases when the 75% to 25%
interval does not match most or all of the discoveries because of the
presence of large outliers, the 95% to 5% interval should be used to
match current ones.
Matching proceeds as follows: Point estimates derived by LDSCV
or NDSCV are used fi rst to predict individual pool sizes because they
generally yield reasonable estimates. In cases when the point estimates
do not predict all discoveries because variances are underestimated
and/or the means are overestimated by LDSCV or NDSCV, the estimated values of μ and σ
2 are chosen from their prediction intervals and
reapplied in the matching process to predict all discovered pool sizes.
The procedure for fi nding μ and σ
2 values is as follows:
Examine which
1.
μ and σ
2 can accommodate most discoveries
with the 75th to 25th or 95th to 5th percentile interval.
Determine the degree of best prediction by measuring the
2.
minimum distance between the discovered pool size and the
median or mean of the predicted pool size.
Verify statistical predictions by examining their geological
3.
implications. After each individual pool-size-by-rank prediction, we can observe whether the implications conform to the
geological model.
Examples of the types of question that one should ask after each prediction are: Have we discovered the largest pool yet? What are the sizes
of the remaining largest pools? What is the potential of the remaining undiscovered pools? Have we predicted enough small pools for the