Hi Sumedha,

@1: there are many different effect sizes which can be transformed into each other. The context in which a specific effect size is used is mainly defined by tradition and ease of computation (e.g. Cohen's d or Hedge's g in two-sample designs, r in correlational designs, eta² is often given in ANOVA designs), but it does not mean that it would be impossible to transform a given effect size into another type of effect size. In your case, it seems to be a bit strange to convert perfectly meaningful path coefficient into a (less easily interpretable) g. Nevertheless you would interpret the *size* of the effect size using some guidelines. I am not sure if there are specific HCI effect size guidelines, but for instance Cohen (1988) has published some guidelines (which are on their own questionable, but in many cases one has to use at least something). 

@2: Standard errors for effect sizes can be computed for any effect size type with slightly different formulas. The authors did not indicate which program they uses (I think the graphs are SPSS, but SPSS does not have any meta-analytic facilities), so I am not completely sure how they computed the SE (typically it is computed by weighting of the variance and sample size).

However, aside of the unclear description of why they transformed the effect sizes and which program they used to conduct their meta-analysis, I am baffled by one other aspect in this meta-analysis as well: If I understand the authors correctly, they simply collect all path coefficients and use those in their analyses. However, path coefficients from different models are only directly comparable if the exactly same set of other predictors ("model") has been used in all original studies. I am not sure that this is indeed the case here.