Effect Strategies

Effect Strategies are being used to compute effect sizes of every outcome variables in an Experiment. Like most methods, we configure an Effect Strategy by specifying the method name, and its appropriate parameters, if any.

Configuration: Effect Strategy

{
    "experiment_parameters": {
        ...,
        "data_strategy": {...},
        "test_strategy": {...},
        "effect_strategy": {
            "name": "MethodName"
        }
    }
}

Mean Differences

• TODO: Add descsription and reference

$$d = \mu_1 - \mu_2$$

"effect_strategy": {
    "name": "MeanDifference"
}

Standardized Mean Differences

• TODO: Add descsription and reference

$$d = \frac{\bar{x}_1 - \bar{x}_2}{s} = \frac{\mu_1 - \mu_2}{s}$$

where $s$ is an average standard deviation of both groups, $\sqrt{\frac{s_{1}^2 + s_{2}^2}{2}}$, where the variance for one of the groups is defined as

$$s_i^2 = \frac 1 {n_1-1} \sum_{i=1}^{n_1} (x_{1,i} - \bar{x}_1)^2$$

"effect_strategy": {
    "name": "StandardizedMeanDifference"
}

Cohen's D

• TODO: Add descsription and reference

$$d = \frac{\bar{x}_1 - \bar{x}_2}{s} = \frac{\mu_1 - \mu_2}{s}$$

where $s$, the pooled standard deviation is defined as:

$$s = \sqrt{\frac{(n_1-1)s^2_1 + (n_2-1)s^2_2}{n_1+n_2 - 2}}$$

where the variance for one of the groups is defined as

$$s_i^2 = \frac 1 {n_1-1} \sum_{i=1}^{n_1} (x_{1,i} - \bar{x}_1)^2$$

"effect_strategy": {
    "name": "CohensD"
}

Hedge's G

• TODO: Add descsription and reference

"effect_strategy": {
    "name": "HedgesG"
}

---

Last update: 2021-09-18