Arboriculture & Urban Forestry 35(5): September 2009 1994). The effect size parameter (d) was calculated with (s), the pooled standard deviation of the means and (J) a weight- ing term that approaches one as sample size increases (Cooper and Hodges 1994) (Equation 3). The weighting factor (J) was calculated with treatment replication, nt and control replication, rameter (d) is very important as it corrects for an overestimation bias when sample sizes are small (Cooper and Hodges 1994), and d can be used for statistical tests of unequal sample sizes (Hunter and Schmidt 2004). Effect size will increase with in- creasing % change, decreasing variance, and increasing sample size. Consequently, effect size values close to zero (i.e., -0.2 to 0.2) relate weak responses relative to values farther from zero. nc (Cooper and Hodges 1994) (Equation 2). The effect size pa- % change (Δ) = [(Xt – Xc ) / Xc ] * 100 Weighting factor (J) = 1 – (3 / (4 * (nt + nc Effect size (d) = (Δ / s) * J - 2) – 1) Treatment effects (Δ and d) were coded as positive or negative according to their interpreted impact on tree, soil, or environmen- tal quality. For instance, a significant decrease in bulk density due to an organic treatment was assigned positive Δ and d values, even though the observed treatment response was a decrease rela- tive to the control. Treatment effects were only calculated for data showing significant (p ≤ 0.05) response on at least half of the data presented. For example, if soil temperature under mulch was sig- nificantly less than under bare ground on six of the ten measured dates, then treatment effects were quantified. Conversely, if soil pH was only significantly less under mulch compared to turf at one of the six measured depths, treatment effects were not quanti- fied. The 33 studies in the meta-analysis spanned many different tree species, growing conditions, soil types, organic treatments, controls, and potential treatment interactions (Table 1). Details on species, soil characteristics, and other specifications (ages, care, potential interactions, etc.) were used for data interpretation. Meta-analysis class variables (i.e., attribute categories) were established to lump significant responses into ecologically het- erogeneous groups (Lipsey 1994). All significant responses were coded into the following seven attribute categories: 1) shoot growth, 2) root growth, 3) physiological, 4) soil chemical, 5) soil physical, 6) soil biological, and 7) environmental (Table 2). Treatment groups were identified according to the type of organic material (mulch, compost, and mix of mulch and com- post) and mode of application [deep surface > 10 cm (4 in), sur- face 0 to 10 cm (0 to 4 in), and applied as in backfill] (Table 1). Statistical tests were performed to identify differences relating to the type of organic material (mulch, compost, or mix) and mode of application (surface, deep surface, or backfill). The interaction between type and mode of application was not significant for any of seven attribute categories (p ≥ 0.844). However, the availability of data likely limited our ability to adequately test for this interac- tion. No studies reported data for mulch as backfill or compost as a deep surface application. Only two studies used mixed materials as backfill. Only three studies reported data for compost applied to the surface and mixed materials on the surface. Other statistical tests were performed, such as the effect of experimental realm (e.g., field versus container environment), but these tests did not reveal 223 any significant (p ≥ 0.270) differences in field versus container studies for the entire data set or for any of the attribute categories. Frequency distributions were compiled and data normality was tested with the Shapiro-Wilk test (SAS 2005). Analysis of variance with Tukey-Kramer HSD and Student’s t-test were used to identify significant differences for various statistical tests (SAS 2005). All statistical differences are reported at the p ≤ 0.05 probability level. RESULTS In total, 79 significant tree, soil, and environmental responses were identified in 33 studies (Figure 1). The distribution for percentage change (Δ) was nonnormal (W = 0.778; p < 0.0001) and heavily weighted (69%) in the 0% to 100% response rela- tive to control range (Figure 1). The mean Δ for all significant responses was 44.8 (SE±9.1), and this value was significantly (p < 0.0001) greater than a null hypothesis, zero response (Fig- ure 1). The effect size (d) parameter removed the bias associ- ated with small sample sizes. The distribution of d was also nonnormal (W = 0.581; p < 0.0001) and heavily weighted (83%) in the zero to one response range (data not depicted). The mean d value for all responses was 0.58 (SE±0.2); also significantly greater than a null response of zero (p = 0.0034). Figure 1. Frequency distribution of % response relative to control (Δ) for seventy-nine significant tree, soil, and environmental attri- bute responses to organic materials, detected in 33 studies. The distribution of significant responses among attribute cat- egories was: shoot growth (18), root growth (9), physiological (20), soil chemical (5), soil physical (18), soil biological (4), and environmental (5) (Figure 2; Table 2). Significant differences were identified for both Δ and d across these attribute categories (Figure 2). Percent response relative to control was significantly (p = 0.0028) greater for soil biological (160.0) compared to envi- ronmental (7.2), soil chemical (10.8), physiological (25.2), shoot growth (34.7), and soil physical (36.8) (Figure 2). Although not significant, root growth Δ (113.7) was less than soil biological and greater than other attribute categories (Figure 2). The d value was significantly (p = 0.0444) greater for soil physical (2.10) com- pared to soil biological (0.00), environmental (0.00), physiologi- ©2009 International Society of Arboriculture
September 2009
| Title Name |
Pages |
Delete |
Url |
| Empty |
Search Text Block
Page #page_num
#doc_title
Hi $receivername|$receiveremail,
$sendername|$senderemail wrote these comments for you:
$message
$sendername|$senderemail would like for you to view the following digital edition.
Please click on the page below to be directed to the digital edition:
$thumbnail$pagenum
$link$pagenum
Your form submission was a success. You will be contacted by Washington Gas with follow-up information regarding your request.
This process might take longer please wait
