How to use the pymer4.stats.discrete_inverse_logit function in pymer4

blups = np.array(
        [np.random.normal(est, sigma, num_grps) for est, sigma in zip(b, grp_sigmas)]
    ).T

    # Generate data
    for grp in range(blups.shape[0]):
        # Create a random design matrix per group
        if corrs:
            x = easy_multivariate_normal(num_obs, num_coef, corrs, mus, sigmas, seed)
        else:
            x = np.random.normal(mus, sigmas, size=(num_obs, num_coef))
        x = np.column_stack([np.ones((num_obs, 1)), x])
        # Use blups to generate group data
        y = np.dot(x, blups[grp, :]) + np.random.normal(*noise_params, size=num_obs)
        if family == "binomial":
            y = discrete_inverse_logit(y)
        if grp == 0:
            x_all, y_all = x, y
        else:
            y_all = np.append(y_all, y, axis=0)
            x_all = np.append(x_all, x, axis=0)

    grp_ids = np.array([[elem] * num_obs for elem in range(1, num_grps + 1)]).ravel()

    data = pd.DataFrame(
        np.column_stack([y_all, x_all[:, 1:], grp_ids]),
        columns=["DV"]
        + ["IV" + str(elem + 1) for elem in range(x_all.shape[1] - 1)]
        + ["Group"],
    )
    blups = pd.DataFrame(
        blups,

assert (
        isinstance(noise_params, tuple) and len(noise_params) == 2
    ), "noise_params should be a tuple of (mean,std)"

    # Generate random design matrix
    if corrs is not None:
        X = easy_multivariate_normal(num_obs, num_coef, corrs, mus, sigmas, seed)
    else:
        X = np.random.normal(mus, sigmas, size=(num_obs, num_coef))
    # Add intercept
    X = np.column_stack([np.ones((num_obs, 1)), X])
    # Generate data
    Y = np.dot(X, b) + np.random.normal(*noise_params, size=num_obs)
    # Apply transform if not linear model
    if family == "binomial":
        Y = discrete_inverse_logit(Y)
    dat = pd.DataFrame(
        np.column_stack([Y, X[:, 1:]]),
        columns=["DV"] + ["IV" + str(elem + 1) for elem in range(X.shape[1] - 1)],
    )

    return dat, b