In terms of non-farm business sales, the integration in the local economy is weak in
Silves and Lixa and very weak in Tavira. On the contrary, it is moderate in Esposende
and strong in Mirandela and Vila Real.
The non-farm business purchase local integration is weak in every town but Tavira,
where it is very weak and in Esposende where it shows values pointing to a moderate
integration.
The employment measured in terms of WTE and the correspondent salary payments
have a high local integration level.
The non-farm businesses employment shows a high local integration with LII values
oscillating between 0,9 and 1.
Unlike the economic footprint analysis, actual transaction valuers are not incorporated
into the variables in this case. Analysis is based directly based on the proportion of the
saçles ans purchases in each zone. On may want to anlyse a more realistic relative
comparison between firms of different size and ownership status. For example, smaller
firms by their very nature are more likely to deal with smaller transaction values, thus
incorporating such values in the analysis may may not psint s realistic picture of the
spatial pattern of these transaction in comparison to large firms.
Independent variables are derived from those already incorporated into the descriptive
analysis, for example related to firm size, firm type, type of the organization, location of
the firm, years in this address, origin of the manager. The bi-variate analysis can be a set
upt to compare one typt to another. For example: independent firms with no sites vs
national and international companies; firms located in uraban areas vs located in rural
areas.
The analysis involves comparing means of the respective dependent variables for the
various independent comparisons of interest. This has to be done in two stages:
- First, compute mean and standard deviation for each comparison. In SPSS
(Analyze/Means/Compare Means);
- Whlist independent comparisons of means is usually done un-paired, t-tests, it
is likely that the dependent variables will noy confirm the assumptions of a
normal distribution and equality fo variances (This is based on previous
experiences with data of this kind. It will be preferable to use the un-paired t-
test if the data will allow),in each case of a Man Whitney U test for 2
independent samples and a Kruskal-Wallis, for more than two samples will
provide a more robust results. This is the non-parametric equivalent of the un-