{p 4 8 2}{cmd:raschpower} allows estimating the power of the Wald test comparing the means of two groups of patients in the context of the Rasch model or the partial-credit model. The estimation is based
on the estimation of the variance of the difference of the means based on the Cramer-Rao bound.
{title:Options}
{p 4 8 2}{cmd:n0} and {cmd:n1} indicates the numbers of individuals in the two groups [100 by default].
{p 4 8 2}{cmd:gamma} indicates the group effect (difference between the two means) [0.5 by default].
{p 4 8 2}{cmd:var} is the expected values of the variances of the latent trait (1 by default): if this option contains only one value, variances are considered as equal between the two groups; if this option contains two values, variances are considered as unequal between the two groups.
{p 4 8 2}{cmdab:d:ifficulties} is a matrix containing the item parameters [one row per item, one column per positive modality - (-1.151, -0.987\-0.615, -0.325\-0.184, -0.043\0.246, 0.554\0.782, 1.724) by default].
{p 4 8 2}{cmd:method}({it:method}) indicates the method for constructing data. ({it:method}) may be GH, MEAN, MEAN+GH or POPULATION+GH [default is method(GH) if number of patterns<500, method(MEAN+GH) if 500<=number of patterns<10000,
method(MEAN) if 10000<=number of patterns<1000000, method(POPULATION+GH) otherwise].
{p 8 14 2} {bf:GH}: The probability of all possible response patterns is estimated by Gauss-Hermite quadratures.
{p 8 14 2} {bf:MEAN}: The mean of the latent trait for each group is used instead of Gauss-Hermite quadratures.
{p 8 14 2} {bf:MEAN+GH}: In a first step, the MEAN method is used to determine the most probable patterns. In a second step, the probability of response patterns is estimated by Gauss-Hermite quadratures on the most probable patterns.
{p 8 14 2} {bf:POPULATION+GH}: The most frequent response patterns are selected from a simulated population of 1,000,000 of individuals. The probability of the selected response patterns is estimated by Gauss-Hermite quadratures.
{p 4 8 2}{cmd:detail} allows a comparison with the classical formula (for manifest variable).
{p 4 8 2}{cmd:expectedpower} allows searching for a sample size in order to reach a fixed level of power (the obtained sample sizes take into account the ratio between $n0$ and $n1$).
{p 4 8 2}Hardouin J.B., Amri S., Feddag M., S<>bille V. (2012) Towards Power And Sample Size Calculations For The Comparison Of Two Groups Of Patients With Item Response Theory Models. Statistics in Medicine, 31(11): 1277-1290.{p_end}
{p 4 8 2}Blanchin M., Hardouin J.B., Guillemin F., Falissard B., S<>bille V. (2013) Power and sample size determination for the group comparison of patient-reported outcomes with Rasch family models. PLoS ONE, 8(2): e57279.{p_end}
{p 4 8 2}Feddag M.L., Blanchin M., Hardouin J.B., S<>bille V. (2014) Power analysis on the time effect for the longitudinal Rasch model. Journal of Applied Measurement: (under press).{p_end}
{p 4 8 2}Guilleux A., Blanchin M., Hardouin J.B., S<>bille V. (2014) Power and sample size determination in the Rasch model: Evaluation of the robustness of a numerical method to non-normality of the latent trait. Plos One, 9(1): e0083652.{p_end}