Effects of Roads on the Structure of Freshwater Turtle Populations

Transcription

1 Effects of Roads on the Structure of Freshwater Turtle Populations DAVID A. STEEN AND JAMES P. GIBBS 350 Illick Hall, 1 Forestry Drive, State University of New York College of Environmental Science and Forestry, Syracuse, NY 13210, U.S.A. Abstract: Road mortality has the potential to alter the structure of turtle populations because turtle populations are highly sensitive to additive sources of adult mortality. To address the issue, we captured painted turtles ( Chrysemys picta; n = 174) and snapping turtles ( Chelydra serpentina; n = 56) in 18 wetlands surrounded by low road density ( 1.5 km roads/km 2 of landscape) and 17 wetlands surrounded by high road density (>1.5 km/km 2 ) in central New York in High road density was associated with male-biased sex ratios in painted turtles (74% vs. 54% males; p = 0.01) and snapping turtles (95% vs. 74% males; p = 0.08), whereas turtle morphology and abundance were not associated with road density. Disproportionate road mortality of females on nesting migrations is the most likely cause of skewed sex ratios. Key Words: Chelydra serpentina, Chrysemys picta, painted turtle, road mortality, roads, sex ratio, snapping turtle, turtles Efectos de Caminos sobre la Estructura de Poblaciones de Tortugas Dulceaucícolas Resumen: La mortalidad en caminos tiene el potencial de alterar a las poblaciones de tortugas porque son relativamente sensibles a causas aditivas de mortalidad de adultos. Para abordar este tema, en 2002 capturamos Chrysemys picta (n = 174) y Chelydra serpentina (n = 56) en 18 humedales con baja densidad de caminos ( 1.5 km caminos/km 2 de paisaje) y en 17 humedales con alta densidad de caminos (>1.5 km caminos/km 2 de paisaje) en New York central. La densidad alta de caminos se asoció con proporciones de sexo sesgadas hacia machos en Chrysemys picta (74% vs. 54% machos; p = 0.01) y en Chelydra serpentina (95% vs. 74% machos; p = 0.08), mientras que la morfología y abundancia de tortugas no estuvieron asociadas con la densidad de caminos. La excesiva mortalidad de hembras en migración reproductiva es la causa más probable del sesgo en las proporciones de sexo. Palabras Clave: Chelydra serpentina, Chrysemys picta, carreteras, mortalidad en caminos, proporción de sexos, tortugas Introduction Turtle conservation may warrant special consideration in relation to roads because turtle life histories are characterized by low annual recruitment rates, high adult survival rates, and delayed sexual maturity (Congdon et al. 1993, 1994). As a consequence, populations have difficulty absorbing the loss of sexually mature individuals (Brooks Address correspondence to J. P. Gibbs, jpgibbs@syr.edu Paper submitted May 28, 2003; revised manuscript accepted December 3, et al.1991). Additionally, the life cycles of many turtle species incorporate terrestrial movements, including annual migrations to nesting sites, migration of juveniles, movement to escape unfavorable habitat conditions, or movement of males to find mates (Gibbons 1986). When such movements intersect with roads, turtles are particularly vulnerable to road-associated mortality because of their relatively slow travel speed. Individual turtles are killed frequently on roads (e.g., Goodman et al. 1994; Ashley & Robinson 1996) because many turtles nest during the hours surrounding dawn and dusk (Legler 1954; Ernst 1986), times that correspond to 1143, Pages

2 1144 Roads and Turtle Populations Steen & Gibbs periods of peak traffic volume (Festin 1996). Although male aquatic turtles occasionally travel overland (Gibbons 1986), sexually mature females completing annual nesting migrations are likely much more susceptible to vehicleinduced mortality than males. Moreover, gravid females of some aquatic species may make multiple excursions overland prior to actual nesting (Christens & Bider 1987; Reese & Welsh 1997). Together, these movements expose turtles, particularly females, to high levels of road mortality ( Haxton 2000). Road networks within the United States are undergoing a steady expansion (National Research Council 1997), but whether this plays a role in the imperilment of turtles in the United States, a country of particularly high turtle diversity and imperilment (Ernst et al. 1994), is unknown (see Gibbons et al. 2000). Gibbs and Shriver (2002) estimated through computer modeling that many species of turtles may be subjected to road mortality that exceeds sustainable levels, but empirical confirmation of their predictions is lacking. Our objective was to conduct field studies to determine whether the density of roads surrounding freshwater wetlands affects the population structure of turtles. To this end, turtle abundance, size, sex ratio, and biomass were contrasted among wetlands in landscapes of high and low road density. Methods Study Area We studied turtle populations in 35 wetlands near Syracuse, New York (43 03 N, E), in three landscapes: (1) areas of high road density, particularly areas adjacent to the New York State Thruway and Interstate 81; (2) areas of low road density at Howland Island Wildlife Management Area, Three Rivers Wildlife Management Area, and other similar protected areas; and (3) landscapes in other mostly rural and suburban areas of intermediate road density. Wetlands were 1 13 ha in area, were dominated by emergent vegetation, and had no overland connections to other bodies of water. Turtle Trapping We trapped turtles during May August of 2002 with baited hoop nets (diameter 1 m, mesh 2.5 cm). We deployed three traps at each site and checked each trap daily for a total of 12 trap nights/site. For each turtle captured, we took several measurements, including carapace and plastron length. We then weighed turtles with hand-held scales and sexed them on the basis of external secondary sexual characteristics. In addition to recording the number of leeches on each turtle, we noted the percent cover of algal growth on the carapace. Turtles were notched at the rear of the posterior marginal scutes of the carapace to indicate previous capture (Cagle 1939). To control for sensitivity among scales used, we set the minimum mass of turtles at 0.25 kg during analysis. We released all individuals at point of capture. We designated female and male painted turtles as sexually mature if their plastron lengths exceeded 9.7 cm and 7.0 cm, respectively (Ernst et al. 1994). We designated female snapping turtles as sexually mature if their carapace lengths exceeded 20.0 cm (Mosimann & Bider 1960) and male snapping turtles as such if their plastron length exceeded 14.9 cm (Christiansen & Burken 1979). Occasionally, we found external secondary sexual characteristics an inadequate means of determining the sex of mature snapping turtles in the field. We later determined the sex of these individuals by examining the ratio of precloacal tail length to posterior lobe length (Petokas 1979). Habitat Analysis We characterized each wetland where we trapped turtles based on digital, 1-m-resolution orthophotographs taken between 1994 and 1999 (New York State Interactive Mapping Gateway 2002). This enabled us to estimate the extent of open water, submerged aquatic vegetation, emergent vegetation, scrub/shrub, live timber, and dead timber within each wetland s basin. Water conductivity (to ±0.25%) and ph (to ±0.002 units) was measured in the field with a portable probe. The landscape surrounding each wetland was characterized within a 1-km radius of each wetland sampled. We estimated road density from the total length of roads within 1 km of a trapping site based on U.S. Bureau of the Census Tiger Files (1999). We selected a 1-km-wide buffer because it is likely an appropriate size to incorporate most nesting and dispersal migrations undertaken by both snapping and painted turtles (Whillans & Crossman 1977; Obbard & Brooks 1980; Congdon et al. 1987). We obtained land-use classifications from National Land Cover Data (NLCD), which were derived primarily from Landsat thematic mapper sensor data with a 30-m resolution (EPA 2002). Land-use classifications were condensed into six categories: open water, wetland, forest, low-intensity residential, developed, and agriculture and grasses. Statistical Analysis We divided wetlands into two categories based on median road density within the 1-km buffer among all wetlands studied: those surrounded by km/km 2 of roads (n = 18) versus those surrounded by >1.5 km/km 2 of roads (n = 17). Based on estimated road densities at 500 randomly selected points, these low and high road-density thresholds corresponded to road densities in 62% versus 38% of the landscape, respectively. To determine the abundance of turtles, we converted the number of captures per site to the number of captures per 100 trap nights (recaptured turtles not included). To normalize

3 Steen & Gibbs Roads and Turtle Populations 1145 Table 1. Sex ratio of sexually mature turtles within 35 wetlands sampled in relation to road density in central New York, Low road density High road density males males Species n (%) p a n (%) p a p b Painted turtle Snapping turtle a Probability that sex ratio (males:females) within a wetland road density class was 1:1; Fisher s exact test. b Probability that sex ratio (males:females) did not differ between wetlands with low road density and those with high road density; Fisher s exact test. data, land-use characterizations that were proportions were arcsine transformed and morphological parameters were log transformed prior to analysis. To contrast the ratio of male to female adults among wetland road-density classes, we used a Fisher s exact test (Zar 1984), and we used the same test to determine whether sex ratios differed from parity. Because all other population and habitat parameters were parametric, we contrasted them between road-density classes with a two-sample t test (Zar 1984; SAS Systems, version 8.0, Cary, North Carolina), with equality of variances as determined by an F test (Zar 1984). Results We captured 174 painted turtles and 56 snapping turtles in the 35 wetlands. Of these captures, 157 of the painted turtles and 41 of the snapping turtles were sexually mature. The sex ratio (males:females) in either species did not differ from parity in wetlands at low road density (Table 1). In wetlands at high road density, sex ratios of mature individuals were skewed in favor of males for snapping turtles (21:1) and painted turtles (3:1; Table 1). There was no difference in the abundance of male, female, or total painted turtles (Table 2) or in that of snapping turtles (Table 3). Morphological parameters of the turtles also were not different, with the exception of algal cover. Female painted turtles in wetlands surrounded by a low density of roads had a higher percentage of algal cover than those in wetlands surrounded by a high density of roads (Table 3). In relation to land use surrounding wetlands (Table 4), wetlands surrounded by a low density of roads had a higher percentage of forest cover and a lower percentage of land designated as agriculture or cultivated grasses than did wetlands at low road density (Table 4). Parameters intrinsic to wetlands did not vary in relation to road density, with the exception of water conductivity, which was higher (mean = 344 µs/cm) at wetlands of higher than at those of lower road densities (mean = 124 µs/cm; Table 4). Table 2. Morphological parameters and abundance of painted turtles in relation to road density at 35 wetlands in central New York, Mean ± SD (n a ) Variable low road density high road density t p b Captures ± 61 (18) ± (17) Females captures of sexually mature individuals ± (18) 9.31 ± (17) carapace length (cm) ± 0.45 (46) ± 0.64 (27) plastron length (cm) ± 0.44 (46) ± 0.61 (27) carapace width (cm) ± 0.27 (46) 9.7 ± 0.4 (27) mass (kg) 0.46 ± 0.12 (39) 0.51 ± 0.14 (19) condition (kg/cm) ± (25) ± (13) carapace algal cover (%) 3.20 ± 0.85 (46) 0.74 ± 0.51 (27) leech (total) 0.72 ± 0.16 (46) 1.26 ± 0.75 (27) Males captures of sexually mature individuals ± (18) ± (17) carapace length (cm) ± 0.3 (45) ± 0.2 (56) plastron length (cm) ± 0.27 (45) ± 0.19 (56) carapace width (cm) 9.03 ± 0.18 (45) 9.22 ± 0.14 (56) mass (kg) 0.29 ± 0.06 (45) 0.31 ± 0.06 (56) condition (kg/cm) 0.02 ± 0.00 (32) 0.02 ± 0.00 (39) carapace algal cover (%) 4.61 ± 2.11 (44) 1.05 ± 0.38 (56) leech (total) 1.38 ± 0.70 (45) 0.5 ± 0.15 (56) a Sample size refers to wetlands for capture-frequency variables (18 vs. 17) and to number of individual turtles for all other variables. b Two sample t test for equal or unequal variances, as appropriate.

4 1146 Roads and Turtle Populations Steen & Gibbs Table 3. Morphological parameters and abundance of snapping turtles in relation to road density at 35 wetlands in central New York, Mean ± SD (n a ) Variable low road density high road density t p b Captures ± 33.33(18) ± 21.56(17) Females captures of sexually mature individuals 2.31 ± 4.79 (18) 0.49 ± 2.02 (17) carapace length (cm) ± 2.46 (11) ± 1.41 (8) plastron length (cm) 17.9 ±1.75 (11) ± 1.38 (6) carapace width (cm) ± 2.21 (11) ± 1.10 (8) mass (kg) 5.29 ± 3.54 (11) 4.34 ± 1.84 (8) condition (kg/cm) 0.16 ± 0.04 (9) 0.15 ± 0.03 (7) carapace algal cover (%) 78.6 ± 8.22 (10) ± (8) leech (total) 2.45 ± 1.80 (11) 6.00 ± 4.23 (7) Males captures of sexually mature individuals 6.48 ± 7.86 (18) ± 14.29(17) carapace length (cm) ± 1.14 (14) ± 1.09 (22) plastron length (cm) 22.1 ± 0.8 (12) ± 0.84 (18) carapace width (cm) ± 0.90 (14) ± 0.84 (22) mass (kg) 7.88 ± 2.83 (12) 6.51 ± 2.12 (17) condition (kg/cm) 0.24 ± 0.06 (10) 0.21 ± 0.05 (16) carapace algal cover (%) ± 6.86 (13) ± 6.90 (21) leech (total) 5.79 ± 3.09 (14) 2.18 ± 0.76 (22) a Sample size refers to wetlands for capture-frequency variables (18 vs. 17) and to number of individual turtles for all other variables. b Two sample t test for equal or unequal variances, as appropriate. Discussion The increased proportion of male turtles within wetlands surrounded by a high density of roads suggests that females were killed on roads to the extent that the population structure of freshwater turtles was altered. Although vehicle-induced mortality tends not to discriminate with respect to the age or sex of individual organisms in many species (Bangs et al. 1989), mature female turtles are likely most susceptible as a result of their higher frequency of terrestrial movements associated with annual nesting migrations. The long-term implications of an alteration of sex ratio are unknown. Brooks et al. (1991) noted that turtles Table 4. Land cover within a 1-km radius of wetlands and characteristics of 35 wetlands where turtle populations were sampled in central New York, 2002 (mean ± SD). Low road density High road density (n = 18) (n = 17) t p Landscape open water (%) 4.50 ± ± wetland (%) 5.70 ± ± forest (%) ± ± low-impact residential (%) 0.33 ± ± developed (%) 0.18 ± ± agriculture and grasses (%) 7.40 ± ± Wetland basin perimeter (m) ± ± basin area (ha) 8.80 ± ± open water (ha) 6.61 ± ± submergent vegetation (ha) 2.33 ± ± emergent vegetation (ha) 1.05 ± ± scrub/shrub (ha) 0.23 ± ± living timber (ha) 0.03 ± ± dead timber (ha) 0.09 ± ± ph 7.92 ± ± conductivity (µs/cm) ± ± Two sample t test for equal or unequal variances, as appropriate.

5 Steen & Gibbs Roads and Turtle Populations 1147 are highly susceptible to population declines following an increase in the mortality of reproductive adults. Because of the life-history traits of turtles, populations may persist for decades before the response to disturbance is discernible in populations (Findlay & Bourdages 2000). Many of the roads we incorporated in our study were constructed within the last 30 years, and given the long lives of turtles (Congdon et al. 1993, 1994) it is unlikely that the effects of roads on turtle populations have been fully manifested over this period. Therefore, the skewed sex ratios we identified may indicate incipient changes in turtle populations being brought about by roads. Moreover, because male turtles undertake long-distance migrations to find mates (Morreale et al. 1984; Gibbons 1986), decreasing the number of sexually mature females may eventually increase the probability that males will disperse overland to other aquatic habitats, thereby further reducing the number of adult individuals within a population. The sex ratios we observed are not necessarily accurate portrayals of the turtle populations (Gibbons 1990). We used baited hoop traps, a method that may be male-biased (Ream & Ream 1966). In addition, for many species of turtles, males and females mature at different sizes; this difference in maturation rate may influence perceived and actual sex ratios (Gibbons 1990; Lovich 1996). Because these biases were present at all study sites, however, we believe that the variation in sex ratio observed among populations reflects relative changes in population structure. It is possible that factors other than road-associated mortality could account for the skewed sex ratios observed. Humans may remove individuals from populations opportunistically or through commercial harvests, activities that may be associated with greater access to wetlands proximal to roads. Female turtles on nesting migrations may also be more susceptible to predation in areas with a high density of roads (Wilcove 1985, Oehler & Litvaitis 1996). Predators of turtles, such as raccoons (Procyon lotor; Christiansen & Gallaway 1984) and domestic dogs (Canis familiaris) often reach their highest densities in fragmented landscapes characterized by suburban development (Hoffman & Gottschang 1977; Wilcove 1985); however, this land-use type had minor representation around the wetlands we sampled (Table 4). Variation in land use surrounding wetlands also could influence sex ratios in populations via temperaturedependent sex determination of developing embryos (Ewert & Nelson 1991; Janzen 1994). In both painted and snapping turtles, warmer nest temperatures typically produce females, whereas cooler temperatures produce males, with the exception that extreme low temperatures in snapping turtles also produce females (Schwartzkopf & Brooks 1985; Ewert & Nelson 1991). However, warm roadsides (Asaeda & Ca 1993), which provide nesting sites for aquatic turtles along with more open lands surrounding roads (Table 4), would be expected to bias the sex ratio of turtle populations toward females in wetlands surrounded by a high road density, not toward males as we observed. Similarly, some contaminants or chemicals associated with agriculture may feminize turtle embryos (e.g., Bergeron et al. 1994), but given the greater extent of agriculture near roaded wetlands (Table 4), we would expect an increased proportion of female, not male, turtles within associated populations. Our findings indicate that roads may be altering the population structure of freshwater turtles. Sex-ratio skew may be indicative of incipient population declines with significant implications for the future status of populations of these long-lived organisms. We conclude that the density of roads surrounding wetlands should be taken into account in efforts to conserve populations of freshwater turtles. Acknowledgments Funding for this study was provided by the Upstate Herpetological Association, the Edna Bailey Sussman Fund, the Maurice Alexander Wetlands Research Award, and the Roosevelt Wild Life Station at the State University of New York College of Environmental Science and Forestry. A. Knox provided field assistance. Comments on the study and manuscript were provided by G. Baldassarre, A. Breisch, P. Ducey, K. Godwin, members of the Hairy Wart Biter Laboratory, and two anonymous reviewers. Permission to access study wetlands was provided by the New York State Thruway Authority, the New York State Department of Transportation, the New York State Department of Environmental Conservation, the Finger Lakes National Forest, the State University of New York College of Environmental Science and Forestry and the Save the County Land Trust. Literature Cited Asaeda, T., and V. A. Ca The subsurface transport of heat and moisture and its effect on the environment: a numerical model. Boundary Layer Meteorology 65: Ashley, E. P., and J. T. Robinson Road mortality of amphibians, reptiles and other wildlife on the Long Point Causeway, Lake Erie, Ontario. Canadian Field Naturalist 110: Bangs, E. E., T. N. Bailey, and M. F. Portner Survival rates of adult female moose on the Kenai Peninsula, Alaska. Journal of Wildlife Management 53: Bergeron, J. M., D. Crews, and J. A. McLachlan PCBs as environmental estrogens: turtle sex determination as a biomarker of environmental contamination. Environmental Health Perspectives 102: Brooks, R. J., G. P. Brown, and D. A.Galbraith Effects of a sudden increase in natural mortality of adults on a population of the common snapping turtle (Chelydra serpentina). Canadian Journal of Zoology 69: Cagle, F. R A system of marking turtles for future identification. Copeia 1939: Christens, E., and J. R. Bider Nesting activity and hatching success

6 1148 Roads and Turtle Populations Steen & Gibbs of the painted turtle (Chrysemys picta marginata) in southwestern Quebec. Herpetologica 43: Christiansen, J. L., and R. R. Burken Growth and maturity of the snapping turtle (Chelydra serpentina) in Iowa. Herpetologica 35: Christiansen, J. L., and B. J. Gallaway Raccoon removal, nesting success, and hatchling emergence in Iowa turtles with special reference to Kinosternon flavescens (Kinosternidae). The Southwestern Naturalist 29: Congdon, J. D., G. L. Breitenbach, R. C. van Loben Sels, and D. W. Tinkle Reproduction and nesting ecology of snapping turtles (Chelydra serpentina) in southeastern Michigan. Herpetologica 43: Congdon, J. D., A. E. Dunham, and R. C. Van Loben Sels Delayed sexual maturity and demographics of Blanding s turtles (Emydoidea blandingii): implications for conservation and management of longlived organisms. 7: Congdon, J. D., A. E. Dunham, and R. C. Van Loben Sels Demographics of common snapping turtles (Chelydra serpentina): implications for conservation and management of long-lived organisms. American Zoologist 34: Ernst, C. H Ecology of the turtle, Sternotherus odoratus, in southeastern Pennsylvania. Journal of Herpetology 20: Ernst, C. H.; J. E. Lovich, R. W. Barbour Turtles of the United States and Canada. Smithsonian Institution Press, Washington, D.C. Ewert, M. A., and C. E. Nelson Sex determination in turtles: diverse patterns and some possible adaptive values. Copeia 1: Festin, S. M., Summary of national and regional travel trends: Office of Highway Information Management, Federal Highway Administration, U.S. Department of Transportation, Washington, D.C. Available from bluebook.pdf (accessed February 2002). Findlay, C. S., and J. Bourdages Response time of wetland biodiversity to road construction on adjacent lands. 14: Gibbons, J. W Movement patterns among turtle populations: applicability to management of the desert tortoise. Herpetologica 42: Gibbons, J.W Sex ratios and their significance among turtle populations. Pages in J. W. Gibbons, editor. Life history and ecology of the slider turtle. Smithsonian Institution Press, Washington, D.C. Gibbons, J. W., D. E. Scott, T. R. Ryan, K. A. Buhlmann, T. D. Tuberville, B. S. Metts, J. D. Greene, T. Mills, Y. Leiden, S. Poppy, and C. T. Winne The global decline of reptiles, déjà vu amphibians. BioScience 50: Gibbs, J. P., and W. G. Shriver Estimating the effects of road mortality on turtle populations. 16: Goodman, S. M., M. Pigeon, and S. O Connor Mass mortality of Madagascar radiated tortoises caused by road construction. Oryx 28: Haxton, T Road mortality of snapping turtles, Chelydra serpentina, in central Ontario during their nesting period. Canadian Field Naturalist 114: Hoffman, C. O., and J. L. Gottschang Numbers, distribution, and movements of a raccoon population in a suburban residential community. Journal of Mammalogy 58: Janzen, F. J Vegetational cover predicts the sex ratio of hatchling turtles in natural nests. Ecology 75: Legler, J. M Nesting habits of the Western Painted Turtle, Chrysemys picta bellii (Gray). Herpetologica 10: Lovich, J. E Possible demographic and ecologic consequences of sex ratio manipulation in turtles. Chelonian Conservation and Biology 2: Morreale, S. J., J. W. Gibbons, and J. D. Congdon Significance of activity and movement in the yellow-bellied slider turtle (Pseudemys scripta). Canadian Journal of Zoology 62: Mosimann, J. E., and J. R. Bider Variation, sexual dimorphism, and maturity in a Quebec population of the common snapping turtle, Chelydra serpentina. Canadian Journal of Zoology 38: National Research Council Towards a sustainable future: addressing the long-term effects of motor vehicle transportation on climate and ecology. National Academy Press, Washington, D.C. New York State Interactive Mapping Gateway New York State 1 meter statewide digital orthophotoimagery New York State GIS Clearinghouse, Albany. Available from nysgis.state.ny.us/gateway/mg/cir.htm (accessed February 2002). Obbard, M. E., and R. J. Brooks Nesting migrations of the snapping turtle (Chelydra serpentina). Herpetologica 36: Oehler, J. D., and J. A. Litvaitis The role of spatial scale in understanding responses of medium-sized carnivores to forest fragmentation. Canadian Journal of Zoology 74: Petokas, P. J The turtles of Cranberry Creek Marsh: population and morphological parameters. M.S. thesis, State University of New York College of Environmental Science and Forestry, Syracuse. Ream, C., and R. Ream The influence of sampling methods on the estimation of population structure in painted turtles. American Midland Naturalist 75: Reese, D. A., and H. H. Welsh Use of terrestrial habitat by western pond turtles, Clemmys marmorata: implications for management. Pages in Proceedings an international conference conservation, restoration, and management of tortoises and turtles: an international conference. New York Turtle and Tortoise Society, Purchase. Schwartzkopf, L., and R. J. Brooks Sex determination in northern painted turtles: effect of incubation at constant and fluctuating temperatures. Canadian Journal of Zoology 63: U.S. Environmental Protection Agency Multi-resolution land characteristics consortium. U.S. Environmental Protection Agency, Washington, D.C. Available from (accessed February 2002). Whillans, T. H., and E. J. Crossman Morphological parameters and spring activities in a central Ontario population of midland painted turtle, Chrysemys picta marginata (Agassiz). Canadian Field-Naturalist 91: Wilcove, D. S Nest predation in forest tracts and the decline of migratory songbirds. Ecology 66: Zar, J. H Biostatistical analysis. Prentice-Hall, Englewood Cliffs, New Jersey.