90
Sampling Design and Statistical Inference for Ecological Assessment
upper and lower ends were sampled in the spring
to measure concentration of chemical constituents
related to acidification, and these data provide a
baseline against which surface-water quality trends
and patterns can be measured through future sampling.
6.11 Conclusion
Sampling design begins with a statement of the research questions and study objectives. With the research questions in hand, a population and variables of interest can be identified. The population
and variables should be relevant with respect to the
study objectives, and they should be appropriate for
addressing the research questions through statistical inference. One of the foremost difficulties in
applying statistical inference for ecological assessment is that of defining the population of interest.
Usually, limitations must be imposed on the original study objectives, because the population must
be a precisely defined collection of units that can
be sampled. The NSS example shows how a complex set of objectives regarding acidification of
streams across large regions and over time can be
addressed. The pivotal idea was that of identifying
the population of interest to be a finite set of stream
reaches covering the region. Then it was straightforward, but clever nonetheless, to select a stratified random sample by randomly positioning a lattice over each subregion map and selecting each
reach located under a lattice point. Resources such
as financing, personnel, and material for sampling
must be identified once the population and variables have been determined. Then a sampling design and a methodology for data analysis can be
proposed, given resource constraints. A proposal
for data analysis methods should identify the parameters and estimators, confidence intervals, hypothesis tests, and predictive models to be used. At
this stage, sample size calculations are necessary to
ensure that the estimates and predictions will be
sufficiently accurate and that the tests will be sufficiently sensitive. Sometimes simple sample size
calculations are adequate, but large-scale sampling
designs often necessitate Monte Carlo simulation
studies.
Increasingly larger data sets are being used for
ecosystem assessment as automated data collection
methods become more widespread. A risk posed by
massive data sets is the tendency to ignore statistical validity until data analysis commences, and then
to proceed with data analysis even when data do
not constitute a valid sample. Proper sampling design will produce data that are appropriate for their
intended purpose and ensure that valid inference is
drawn from the data. Statistical validity and proper
sampling design are critical to nearly all quantitative ecological assessment efforts, no matter how
many observations are collected. Researchers must
be vigilant in their efforts to use sampling designs
and expect the same of others.
6.12 References
Austin, M. P.; Heyligers, P. C. 1989. Vegetation survey
design for conservation: gradsect sampling of forests
in north-eastern New South Wales. Bioi. Conserv.
50:13-32.
Bourgeron, P. S.; Humphries, H. c.; Jensen, M. E. 1994.
General sampling considerations for landscape evaluation. In: Bourgeron, P.S.; Jensen, M.E., eds. Volume
II: ecosystem management: principles and applications. Portland, OR: U.S. Dept. Agric., For. Serv., Pacific Northw. Res. Sta.: 109-120.
Cooper, S. V.; Neiman, K. E.; Steele, R; Roberts, D. W.
1987. Forest habitats of northern Idaho: a second approximation. Ogden, UT: U.S. Dept. Agric., For.
Serv., Intermountain Res. Sta.
Eberhart, L. L.; Thomas, J. M. 1991. Designing environmental field studies. Ecol. Monogr. 61:53-73.
Hansen, M. H.; Hurwitz, W. N.; Madow, W. G. 1953.
Sample survey methods and theory, volume 1: methods and applications. New York: John Wiley & Sons.
Hayne, D. W. 1949. An examination of the strip census
method for estimating animal populations. J. Wildlife
Manage. 13:145-157.
Horvitz, D. G.; Thompson, D. J. 1952. A generalization
of sampling without replacement from a finite universe. J. Amer. Statist. Assoc. 47:663-685.
Kaufmann, P. R; Herlihy, A. T.; Elwood, J. W.; Mitch,
M. E.; Overton, W. S.; Sale, M. J.; Messer, J. J.;
Cougan, K. A.; Peck, D. V.; Reckhow, K. H.; Kinney,
A J.; Christie, S. J.; Brown, D. D.; Hagley, C. A;
Jager, H. I. 1988. Chemical characteristics of streams
in the Mid-Atlantic and southeastern United States,
volume I: population descriptions and physio-chemical
relationships. EPA/600/3-88/021a. Washington, DC:
U.S. Env. Protect. Agency.
Kendall, K. c.; Metzgar, L. H.; Patterson, D. A; Steele,
B. M. 1992. Power of sign surveys to monitor population trends. Ecol. Appl. 2:422-430.
Lesica, P.; Steele, B. M. 1996. A method for monitoring
long-term trends: an example using rare arctic-alpine
plants. Ecol. Appl. 6:879-887.
Levy, P. S.; Lemeshow, S. 1991. Sampling of populations. New York: John Wiley & Sons.
Mueller-Dombois, D.; Ellenberg, H. 1974. Aims and
methods of vegetation ecology. New York: John Wiley & Sons.
Ostle, B.; Mensing, R W. 1975. Statistics in research,
3rd ed. Ames, IA: Iowa State University Press.
Sampling Design and Statistical Inference for Ecological Assessment
upper and lower ends were sampled in the spring
to measure concentration of chemical constituents
related to acidification, and these data provide a
baseline against which surface-water quality trends
and patterns can be measured through future sampling.
6.11 Conclusion
Sampling design begins with a statement of the research questions and study objectives. With the research questions in hand, a population and variables of interest can be identified. The population
and variables should be relevant with respect to the
study objectives, and they should be appropriate for
addressing the research questions through statistical inference. One of the foremost difficulties in
applying statistical inference for ecological assessment is that of defining the population of interest.
Usually, limitations must be imposed on the original study objectives, because the population must
be a precisely defined collection of units that can
be sampled. The NSS example shows how a complex set of objectives regarding acidification of
streams across large regions and over time can be
addressed. The pivotal idea was that of identifying
the population of interest to be a finite set of stream
reaches covering the region. Then it was straightforward, but clever nonetheless, to select a stratified random sample by randomly positioning a lattice over each subregion map and selecting each
reach located under a lattice point. Resources such
as financing, personnel, and material for sampling
must be identified once the population and variables have been determined. Then a sampling design and a methodology for data analysis can be
proposed, given resource constraints. A proposal
for data analysis methods should identify the parameters and estimators, confidence intervals, hypothesis tests, and predictive models to be used. At
this stage, sample size calculations are necessary to
ensure that the estimates and predictions will be
sufficiently accurate and that the tests will be sufficiently sensitive. Sometimes simple sample size
calculations are adequate, but large-scale sampling
designs often necessitate Monte Carlo simulation
studies.
Increasingly larger data sets are being used for
ecosystem assessment as automated data collection
methods become more widespread. A risk posed by
massive data sets is the tendency to ignore statistical validity until data analysis commences, and then
to proceed with data analysis even when data do
not constitute a valid sample. Proper sampling design will produce data that are appropriate for their
intended purpose and ensure that valid inference is
drawn from the data. Statistical validity and proper
sampling design are critical to nearly all quantitative ecological assessment efforts, no matter how
many observations are collected. Researchers must
be vigilant in their efforts to use sampling designs
and expect the same of others.
6.12 References
Austin, M. P.; Heyligers, P. C. 1989. Vegetation survey
design for conservation: gradsect sampling of forests
in north-eastern New South Wales. Bioi. Conserv.
50:13-32.
Bourgeron, P. S.; Humphries, H. c.; Jensen, M. E. 1994.
General sampling considerations for landscape evaluation. In: Bourgeron, P.S.; Jensen, M.E., eds. Volume
II: ecosystem management: principles and applications. Portland, OR: U.S. Dept. Agric., For. Serv., Pacific Northw. Res. Sta.: 109-120.
Cooper, S. V.; Neiman, K. E.; Steele, R; Roberts, D. W.
1987. Forest habitats of northern Idaho: a second approximation. Ogden, UT: U.S. Dept. Agric., For.
Serv., Intermountain Res. Sta.
Eberhart, L. L.; Thomas, J. M. 1991. Designing environmental field studies. Ecol. Monogr. 61:53-73.
Hansen, M. H.; Hurwitz, W. N.; Madow, W. G. 1953.
Sample survey methods and theory, volume 1: methods and applications. New York: John Wiley & Sons.
Hayne, D. W. 1949. An examination of the strip census
method for estimating animal populations. J. Wildlife
Manage. 13:145-157.
Horvitz, D. G.; Thompson, D. J. 1952. A generalization
of sampling without replacement from a finite universe. J. Amer. Statist. Assoc. 47:663-685.
Kaufmann, P. R; Herlihy, A. T.; Elwood, J. W.; Mitch,
M. E.; Overton, W. S.; Sale, M. J.; Messer, J. J.;
Cougan, K. A.; Peck, D. V.; Reckhow, K. H.; Kinney,
A J.; Christie, S. J.; Brown, D. D.; Hagley, C. A;
Jager, H. I. 1988. Chemical characteristics of streams
in the Mid-Atlantic and southeastern United States,
volume I: population descriptions and physio-chemical
relationships. EPA/600/3-88/021a. Washington, DC:
U.S. Env. Protect. Agency.
Kendall, K. c.; Metzgar, L. H.; Patterson, D. A; Steele,
B. M. 1992. Power of sign surveys to monitor population trends. Ecol. Appl. 2:422-430.
Lesica, P.; Steele, B. M. 1996. A method for monitoring
long-term trends: an example using rare arctic-alpine
plants. Ecol. Appl. 6:879-887.
Levy, P. S.; Lemeshow, S. 1991. Sampling of populations. New York: John Wiley & Sons.
Mueller-Dombois, D.; Ellenberg, H. 1974. Aims and
methods of vegetation ecology. New York: John Wiley & Sons.
Ostle, B.; Mensing, R W. 1975. Statistics in research,
3rd ed. Ames, IA: Iowa State University Press.