Attitude Nature Stats Attitudes Table

¶ … Attitude Nature

Nature Stats Attitudes

Table 1 (Appendix I) displays the results of a survey of students asked to rank their attitude toward nature on a continuous integer scale of 1-100. The results present a clear picture of these students' attitudes and also inform further question about their outdoor activities in a number of ways. The results in Table 1 support future hypothesis testing because skewness and Kurtosis are close to normal, which indicates both parametric and non-parametric inferential statistics will likely be appropriate to describe and predict correlation between other variables of interest, which is often not the case for samples this small (n=30).

Table 2, "Attitude About Nature" (Appendix II) lists the frequency of students' ranking of their attitudes for nature, against which other variables can be compared to see if the resulting distributions are statistically significant, i.e. occur as much or less often than would occur by chance at whatever degree of "alpha" or risk of Type I vs. Type II error experimenters decide is of interest, usually 0.05 or 0.01. The results demonstrate that no student answered below the value of 5, and since the top score was 100, therefore no student could answer more than that, which bodes well for the normality of data since this reduces the effect of outliers, which shows up in low skewness and Kurtosis near to normal (below). While the distance between the one "100" answer and the next-highest group of two "80" answers suggests the "100" could possibly be an outlier introducing enough skew to demand non-parametric inferential tests, a number of tests easily performed in SPSS will demonstrate the strength to which that sole "100" result disturbs the normality in the rest of the data.

Figure 1 (Appendix III) shows these results displayed in a frequency histogram with normal curve superimposed. This histogram reveals that the results were similar to normal, with the highest frequency occurring around the median, such that mean and median were very close, which Table 1 reveals was only a difference of one percent (Appendix I). Figure 1 shows that the majority of readings surrounded this mean and median cluster on both sides to symmetrical degree, as opposed to appearing grouped highly above or below. The result of this all is a relatively high value for the first standard deviation, which corresponds to meaningful Kurtosis but very little skew -- Table 1 reveals less than .01 skewness, which means the results are highly symmetric (Brown, 2008-2012, n.p.). Negative Kurtosis reflects the concentration of data in these shoulders (DeCarlo, 1997, p. 294), which is supported by two modes one either side of the mean-median (Figure 2, Appendix IV shows this bimodality, not revealed in Figure 1). While this indicates less than normality and warns that parametric inferential statistics should be scrutinized carefully and non-parametrics (CHI-Squared; rank-sum; Wilcoxon's, etc.) run for corroboration in case the 100 value turns out to be an outlier, these results show more symmetry than is often encountered, and bode well for the robustness of inferential relationship between "Attitudes" and other variables should any such relationships arise.