Laserfiche WebLink
Discriminant function analyses were used to determine if Colorado squawf <br />nursery habitats could be predicted by a model of habitat characteristics. A <br />conducted on numerous combinations of up to 42 variables. Models were develop <br />season of each year, for each season for all years, for each year and for all <br />pooled. Variables appearing consistently in discriminant models included back <br />depth. In spring samples, water temperature and the difference between main c <br />backwater temperatures also contributed to the models. Depending on the varia <br />the stepwise function, area, length and volume also contributed to discriminat <br />Although individual models for season and year explained up to 100% of the var <br />Colorado squawfish backwaters and those not used by Colorado squawfish, most i <br />many variables to be practical for field application. Some models yielded as <br />explanation of the variance. <br />Attempts were made to develop a single discriminant model which would co <br />separate habitats used by Colorado squawfish from those not used, and which wo <br />applicable across years and seasons. Variables entered in this model were sel <br />performance in the stepwise discriminant analyses conducted for each season an <br />data across years increased variability of the data set, thus weakening model <br />These general models described between 10% and 51% of the variability between <br />amount of variability accounted for depended on the degree of pooling. Likewi <br />accurately classify backwaters decreased as more data sets were combined for a <br />used to classify backwaters by Colorado squawfish use, these models had classi <br />rates between 0% and 35% for season by year tests. Classification error rates <br />