network_meta_analysis
heterogeneity.network_meta_analysis(treatment, study, mean, sd, n, reference=None, sm='MD')
Pool benchmark results into one coherent method ranking.
Parameters
| Name | Type | Description | Default |
|---|---|---|---|
treatment |
Sequence[str] | Five parallel sequences, one entry per study arm: the method label, the study label (a (benchmark, dataset) block), the mean rank of the method over the metrics, the standard deviation of those ranks, and the number of metrics it rests on. Each study must contribute at least two arms. Arms with a NaN mean, sd or n are dropped. | required |
study |
Sequence[str] | Five parallel sequences, one entry per study arm: the method label, the study label (a (benchmark, dataset) block), the mean rank of the method over the metrics, the standard deviation of those ranks, and the number of metrics it rests on. Each study must contribute at least two arms. Arms with a NaN mean, sd or n are dropped. | required |
mean |
Sequence[str] | Five parallel sequences, one entry per study arm: the method label, the study label (a (benchmark, dataset) block), the mean rank of the method over the metrics, the standard deviation of those ranks, and the number of metrics it rests on. Each study must contribute at least two arms. Arms with a NaN mean, sd or n are dropped. | required |
sd |
Sequence[str] | Five parallel sequences, one entry per study arm: the method label, the study label (a (benchmark, dataset) block), the mean rank of the method over the metrics, the standard deviation of those ranks, and the number of metrics it rests on. Each study must contribute at least two arms. Arms with a NaN mean, sd or n are dropped. | required |
n |
Sequence[str] | Five parallel sequences, one entry per study arm: the method label, the study label (a (benchmark, dataset) block), the mean rank of the method over the metrics, the standard deviation of those ranks, and the number of metrics it rests on. Each study must contribute at least two arms. Arms with a NaN mean, sd or n are dropped. | required |
reference |
str | None | The reference treatment to express effects against. Defaults to netmeta’s own choice (the first treatment) when None. | None |
sm |
str | The summary measure passed to netmeta; “MD” (mean difference) is the right choice for the mean-rank arms. | 'MD' |
Returns
| Type | Description |
|---|---|
| NetworkMetaReport |
Raises
| Type | Description |
|---|---|
| ValueError | If the sequences differ in length, or fewer than two treatments or two studies remain after dropping incomplete arms. |
| RNotAvailableError | If the R toolchain with netmeta is not available. |
| RExecutionError | If the R subprocess fails. |