6. Normalising results for intuitive visual inspection

The study comes down to evaluating the effect on target variables of available choices about online community management. This evaluation is contingent on the "state of the world", represented by parameters which are assumed not under the control of the community manager. For the purpose of the study, cultural variables (globalchattiness, intimacystrength,randomisedchattiness – the last one intended as a measure of community diversity on the introversion vs.extroversion spectrum) are assumed to be parameters.

For intuitive representation, we proceed as follows.

We model two choices that the community manager (or, more correctly her principal) must make:

1. Switching from an engage policy to a both policy, therefore adding onboard.
2. Switching from a newer to a more active priority to allocate her capacity.

The choice of baselines is arbitrary and should not be interpreted as assuming a policy or a priority as "more natural" than the other.

For each value of the parameters vector, we quantify the result of the choice as a t-value. Moreover, if necessary we invert the sign of the t-value so that a positive value means that the choice being modelled brings about a positive change for the community. For example, consider the case of the variable dropouts. If, switching policy from engage to both, the average number of dropouts goes down, the t-value will have a negative sign because

mean ( dropouts | policy = 'both' ) - mean ( dropouts | policy == 'engage') < 0

However, the effect of switching policy on inclusivity is positive: fewer dropouts are a sign of a more inclusive community. In these cases, we represent the effect on the target variable as -t.

Variables for which we operate this conversion are:

1. dropouts. Decreasing it has a positive effect on inclusivity.
2. ms (Gini coefficient on membershipstrength). Decreasing it has a positive effect on inclusivity.
3. nc (Gini coefficient on number of comments). Decreasing it has a positive effect on diversity.
4. mgmteffort (number of comments by the community manager). It is a measure of costs: decreasing it is, all other things being equal, a measure of efficiency.