Running Head: Invisible Leverage in Adoption of Online Social Support Community

17

regression, which was performed with each of the last three hypotheses to make sure a cubic

relationship does not exist if a curvilinear relationship is found.

Two data transformation procedures were performed in advance of these analyses:

centering and winsorizing. First, data on the three network size measures were centered in order

to remove the multicollinearity problem among these three predictors, especially in curvilinear

regression. Correlation coefficients among the three predictors were found high and indicated

the possibility of such problem (von Eye & Schuster, 1998). Second, data on the six adoption

measures were winsorized in order to remove the effect of outliers on curve estimation. Outliers

were identified by residual analysis, and outliers with the two highest residual outlier values

were replaced with the most adjacent score (Kokic & Bell, 1994).

Weighting was not performed even though the present study employed a stratified

disproportionate sampling technique.

8

This is because the proportion of mental illness OSSCs

among all general OSSCs population is unknown. The researchers concluded that weighting

mental illness OSSCs using the proportion in the study population cannot make any contribution

to the representativeness of data with the general population proportion being unknown.

For hypothesis 2, each measure of network size was separately entered in a third-order

polynomial regression analysis on each measure of adoption first, then all three measures of

network size were simultaneously entered in a third-order polynomial regression analysis with

each measure in each block in order to examine the overall contribution of each network size

measure on adoption.

*Result *

The descriptive analysis shows that the average number of total postings for each

community was 493.35 (SD = 1240.83) and the average number of registered members was

49.80 (SD = 71.29). It indicates that the average number of posters was 53.25 (SD = 75.57) and

the average number of adopters was 28.98 (SD = 46.59). The average frequency of postings per

adopter was 11.00 (SD = 11.11) and the average duration of adopter was 140.78 (SD = 110.46).

When the major contributor’s participation was excluded, the average frequency of postings per

adopter was 7.63 (SD = 7.77) and the average duration of adopter was112.78 (SD = 93.94). The