154 have artificially higher correlations, R2 Hauer et al.: Local Outcomes of Federal and State Urban & Community Forestry Programs , and significance statis- tics. The scaled variables reduce the concern of artificially higher results. A multiple regression model was used as an a priori test of the relationship among independent variables (indicators) and the dependent variable (ActDiff). The dependent variable was derived from the composite change in the number of communi- ties reported to have urban forestry activity in 2002 from 1997 within each of the 41 states used in this study. The composite value was derived from the number of communities in a state that were within one of the four activity rankings. The scaled depen- dent variable ActDiff = (Act2002/number communities in 2002) – (Act1997/number communities in 1997); whereas Act2002 = number of communities within a state reported to have urban for- estry activity in 2002, and Act1997 = number of communities within a state reported to have urban forestry activity in 1997. The seven independent variables TechFreq, TechAsst, Fi- nAsst, FedGrant, StaGrant, FedMoney, and StaMoney (Table 1; Table 2) were initially hypothesized and tested to explain change on ActDiff and selected based on evidence that technical assis- tance, financial assistance, and program money resources lead to a change in activity (Baugman 1980; Still et al. 1996; Vitosh and Thompson 2000; Bird 2002). The model dates were selected since the PMAS data compilation started in 1997 and 2002 was the questionnaire study year. After testing and refinement of the ini- tial a priori model, exploratory testing of additional independent variables (OtherAgn, EnabLeg, FTE, Agency, Coordin, ProgYear, FundAdeq, Council, and StraPlan) occurred through sequential and stepwise multiple regression techniques (Table 2). These vari- ables where used as anecdotal effect is presumed or hypothesized for putative effect and addressed if staffing levels, agency support, funding adequacy, other agencies involved in state U&CF, coor- dination with U&CF delivery, state U&CF council involvement, strategic planning, enabling legislation, and when the state U&CF was created. The final model was cross-validated using the activ- ity difference in communities between 1997 and 2003, 1997 and 2004, interpreting validation through comparable sign and value of parameters. Ideally, validation occurs with a different popula- tion (e.g., country) or in a more distant time period; however, no data currently exists to do this, this approach is offered as the best available, and often validation modeling is not done in studies. Significance for all tests, except where noted, used an a ≤ 0.05 significance level as evidence to reject a null hypothesis that no increase in urban forestry activity occurred. Indicator selection used an a ≤ 0.25 significance level for initial screening of variables and an a ≤ 0.10 significance level for retention in the final model. Outliers within the multiple regression model were discerned us- ActDiffz FinAsstz TechAsstz TechFreq FedGrant StaGrant FedMoneyz StaMoneyz FTEz z 1 0.023 0.478 (**) 0.340 (*) -0.075 0.322 (*) -0.062 0.250 -0.014 FinAsstz 0.023 1 0.339 (*) -0.011 0.198 0.246 0.355 (*) 0.377 (**) 0.401 (**) TechAsstz 0.478 (**) 0.339 (*) 1 0.121 -0.176 0.218 0.473 (**) 0.494 (**) 0.643 (**) ing the Mahalanobis distance procedure at the <0.001 significance level and none was found (Mertler and Vannatta 2005). Assump- tions of normality, linearity, and homoscedasticity were also met using bivariate plots between independent and dependent vari- ables and a plot of the standardized residuals and standardized predicted values from the final multiple regression model. Ex- amination for multicolliniarity in models used variance inflation factor statistics with a lack of multicolliniarity interpreted read as tics with a lack of multicolliniarity interpreted as the variance in- flation factor <10 (Neter et al. 1990; Mertler and Vannatta 2005). RESULTS Local Urban Forestry Activity Mean local urban forestry activity increased between 1997 and 2002 (Figure 1). The composite mean level increase av- eraged 2.1% annually (t-value = 3.979, n = 49, p < 0.000). A likewise 2.1% annual decrease with the mean number of com- munities rated as inactive or nonparticipatory in local urban forestry programming was found (t-value = -2.491, n = 49, p = 0.016). The PMAS categories of sustained (t-value = 2.244, n = 48, p = 0.029), developmental (t-value = 3.181, n = 49, p = 0.003), and project (t-value = 2.632, n = 49, p = 0.011) dem- onstrated significant increases in U&CF activity (Figure 1). Communities rated within the formative category had no sig- nificant change (t-value = 1.616, n = 48, p = 0.113). There was no significant change in the total number of communities (t-value = 0.862, n = 49, p = 0.393) between 1997 and 2002. Initial a priori Model Initial exploratory modeling found staffing level scaled by com- munity as significant and was included in the initial a priori model. From the full model of eight independent variables, four indicators provided evidence of the change in local program activity (Table 3). Pearson’s correlation coefficients also sug- gest a relationship for three independent variables (TechAsst, TechFreq, and StaGrant) and change in activity over the study period (Table 3). The number of communities receiving tech- nical assistance (TechAsst), frequency of technical assistance types to communities (TechFreq), the amount of state govern- ment money allocated to the state U&CF program (StaMoney), and staffing level (FTE) were selected for further testing (Table 3). State money used with grants (StaGrant), Federal Coop- erative Assistance Challenge Cost-share Grants (FedGrant), the Table 2. Pearson’s correlation coefficients of variables used in the initial a priori model. ActDiffz TechFreq 0.340 (*) -0.011 0.121 1 -0.035 0.099 -0.252 0.125 0.014 FedGrant -0.075 0.198 -0.176 -0.035 1 -0.301 (*) 0.222 -0.076 -0.008 Scaled variable by dividing by number of communities in the state from which the data were derived. * Correlation is significant at the 0.05 probability level (2-tailed). ** Correlation is significant at the 0.01 probability level (2-tailed). ©2011 International Society of Arboriculture StaGrant 0.322 (*) 0.246 0.218 0.099 -0.301 (*) 1 -0.339 (*) 0.389 (**) -0.085 FedMoneyz -0.062 0.355 (*) 0.473 (**) -0.252 0.222 1 -0.339 (*) 0.281 (*) 0.686 (**) StaMoneyz 0.250 0.377 (**) 0.494 (**) 0.125 -0.076 0.389 (**) 0.281 (*) 1 0.616 (**)
July 2011
| 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
