2.3.3 Within-lake regressions
To assess whether the across-lake trends identified using mixed models were observable within individual lakes, we performed linear regressions separately at each lake. For each of our focal response variables (see Table 1), we used the same model formulations from the across-lakes analysis (i.e., the explanatory variables from Table 1 that were selected via AICc) to perform regressions within a lake. We saved the resulting coefficient estimates for each explanatory variable used to predict this focal response. We then plotted the distribution of coefficient estimates for all explanatory variables across all lakes, and we compared the median of these distributions to the mixed effect model coefficient estimates. For each response variable, we only included lakes that had at least 10 years of paired data for the response variable and all selected explanatory variables.