Table A: MATLAB's Statistical Toolbox was used to perform an ANCOVA analysis was used to calculate linear regression fits for the dependency of the coefficient of lift and coefficient of drag on velocity ratio (figure 8). This table summarises the results of testing for the similarity of the linear regressions that fit the individual data sets of the five different models to a linear regression that would fit the data of all models combined. Presented are the deviation of the slopes of the individual fits from the slope of the regression that would fit all data. P-values of p~\textgreater~0.05 indicate that the slope of the individual set is not different from the slope of the regression of all data with a probability of 95%. Table B: P-values (rounded to two decimal places) of the pairwise comparisons of the slopes of the regression lines among the five models. The slopes of the regression lines were tested separately for coefficient of lift and drag of downstoke and upstroke, respectively.

Table A: MATLAB's Statistical Toolbox was used to perform an ANCOVA analysis was used to calculate linear regression fits for the dependency of the coefficient of lift and coefficient of drag on velocity ratio (figure 8). This table summarises the results of testing for the similarity of the linear regressions that fit the individual data sets of the five different models to a linear regression that would fit the data of all models combined. Presented are the deviation of the slopes of the individual fits from the slope of the regression that would fit all data. P-values of p~\textgreater~0.05 indicate that the slope of the individual set is not different from the slope of the regression of all data with a probability of 95%. Table B: P-values (rounded to two decimal places) of the pairwise comparisons of the slopes of the regression lines among the five models. The slopes of the regression lines were tested separately for coefficient of lift and drag of downstoke and upstroke, respectively.

The lift and drag data of the five different aspect ratio wings was stored in a MATLAB structure and then converted to XML. For easiest access use xml2mat (available on https://www.mathworks.com/matlabcentral/fileexchange) to convert back to a MATLAB structure. The MATLAB structure contained 5 structures. Each of these structures contains fields named modelID, data and inertia. The structure data is a 5-D matrix, the structure inertia is a 4-D matrix of size. Both structures, data and inertia, have two fields, L and D, which are 2-D matrices of type double, where each column represents one phase-averaged wingbeat cycle. All force data are given in Newton. The data and inertia matrices are sorted so that each combination of test variables per model (freestream velocity, U, downstroke ratio, DR, stroke amplitude angle, a, sweep amplitude angle, b, and flapping frequency, f) is uniquely identified. The order for the data matrix is: [U,DR,a,b,f]. The order for the inertia matrix is: [DR,a,b,f]. U runs from 1 to 2, corresponding to 5.0 m/s and 7.5 m/s; DR runs from 1 to 3, corresponding to to downstroke ratios of .44, .50, .56; a runs from 1 to 7, corresponding to angles from 15° to 110° in increments of 15°; b runs 1 to 5, corresponding to angles from 15° to 55° in increments of 10°; f runs from 1 to 6, corresponding to frequencies of 2, 4, 6 8, 9, and 10 Hz; Example: Access to lift data of the Eptesicus baseline stroke of model AR35bl at a freestream velocity of 5 m/s: AR35bl.data(1,2,7,5,5).L

The lift and drag data of the five different aspect ratio wings was stored in a MATLAB structure and then converted to XML. For easiest access use xml2mat (available on https://www.mathworks.com/matlabcentral/fileexchange) to convert back to a MATLAB structure. The MATLAB structure contained 5 structures. Each of these structures contains fields named modelID, data and inertia. The structure data is a 5-D matrix, the structure inertia is a 4-D matrix of size. Both structures, data and inertia, have two fields, L and D, which are 2-D matrices of type double, where each column represents one phase-averaged wingbeat cycle. All force data are given in Newton. The data and inertia matrices are sorted so that each combination of test variables per model (freestream velocity, U, downstroke ratio, DR, stroke amplitude angle, a, sweep amplitude angle, b, and flapping frequency, f) is uniquely identified. The order for the data matrix is: [U,DR,a,b,f]. The order for the inertia matrix is: [DR,a,b,f]. U runs from 1 to 2, corresponding to 5.0 m/s and 7.5 m/s; DR runs from 1 to 3, corresponding to to downstroke ratios of .44, .50, .56; a runs from 1 to 7, corresponding to angles from 15° to 110° in increments of 15°; b runs 1 to 5, corresponding to angles from 15° to 55° in increments of 10°; f runs from 1 to 6, corresponding to frequencies of 2, 4, 6 8, 9, and 10 Hz; Example: Access to lift data of the Eptesicus baseline stroke of model AR35bl at a freestream velocity of 5 m/s: AR35bl.data(1,2,7,5,5).L

Prior to statistical analysis the data of coefficients of lift and drag for down- and upstroke, respectively, versus velocity ratio for these four cases was grouped into 15 bins for the five different wings individually. In each bin, all data points with a force coefficient more than ± 1 standard deviation from the mean of the force coefficient in the respective bin were excluded from further analysis. The remaining data points per bin were averaged and subjected to an ANCOVA analysis.

Prior to statistical analysis the data of coefficients of lift and drag for down- and upstroke, respectively, versus velocity ratio for these four cases was grouped into 15 bins for the five different wings individually. In each bin, all data points with a force coefficient more than ± 1 standard deviation from the mean of the force coefficient in the respective bin were excluded from further analysis. The remaining data points per bin were averaged and subjected to an ANCOVA analysis.