Can't complete analysis for star networks

[tommycrock]

With a small network I am getting the following error while setting up the evaluation:

sensitity analysis not possible for Low risk within study bias studiesR returned an error: attempt to set 'colnames' on an object with less than two dimensions In call: `colnames<-`(`*tmp*`, value = c("Q_dbt", "df", "pv_dbt"))

When proceeding through the steps, when I select the rule for RoB, the "within-study bias updated" pop-up appears but the graph doesn't show anything. The same is true for indirectness. When I get to imprecision, it oddly shows as a default of -2. I can change this to a sensible value (e.g. 1) and 'set' which shows the panels and risk for each comparison, but then the proceed button doesn't work and I can't select heterogeneity (although it highlights on hover). I can select incoherence. When I move to the report page it is blank for Within-study bias | Reporting bias | Indirectness.

Setup: long binary data, random effects OR (edit: also tested fixed effects OR, RR & RD), comparisons containing one of the interventions.

Additional info: I have successfully set up comparisons with other data. I can provide the dataset.

[tommycrock]

I'm getting similar with some other networks.

sensitity analysis not possible for Low risk within study bias studiesR returned an error: object 'n' not found In call: eval(expr, envir, enclos)

Resulting in none of the domains being ratable.

Also, on imprecision the clinically important effect size defaulted to -2 (SMD) and when I changed it to 0.05 it said values less than 0.05 and greater than 20 would be considered significant!

This was long, continuous, random SMD. I also tried wide format.

(can't switch to heterogeneity) (proceed from Incoherence takes me to the home page)

[tommycrock]

So, I have realised that the 'errors' relating to sensitivity analysis were indeed because there were no low risk studies in the sample. However, when I artificially put some low risk studies in I still get the above problems in not producing results. [edit] I have realised the problems with results not being produced were related to the data being a star network (or similar). If I create a loop in the data by renaming some of the treatments then results are produced. If I create a star with an extra leg the within study bias, reporting bias and indirectness panels work (they don't with a plain star), however the imprecision, heterogeneity and incoherence do not work. I have tried creating a fake loop on the side with 3 comparisons (two fake treatments), six arms, each with y1, sd100, n{1,2,5} but this does not resolve the problems with imprecision, heterogeneity and incoherence.