Any way to calculate the Composite Reliability (CR) value ?

[TasosR83]

Is there any way included to calculate the CR using this software ?

[jezhumble]

Hi. No, it doesn't calculate CR, but you can get hold of Cronbach's Alpha, Dillon-Goldstein Rho, and the eigenvalues of the first and second principal components - use the unidimensionality() method: https://plspm.readthedocs.io/en/latest/source/plspm.plspm.html#plspm.plspm.Plspm.unidimensionality

[TasosR83]

May I propose the following change to the code (I am not an experienced Github user)
Change the inner_summary.py code lines 33-... to:

        ave = pd.Series(np.NaN, index=path.index, name="ave")
        cr = pd.Series(np.NaN, index=path.index, name="cr")
        communality_aux = []
        num_mvs_in_lv = []
        for lv in list(config.path()):
            communality = outer_model.loc[:, "communality"].loc[config.mvs(lv)] # = lamdas^2
            loading = outer_model.loc[:, "loading"].loc[config.mvs(lv)] # lamdas 
            block_communality.loc[lv] = communality.mean()
            mean_redundancy.loc[lv] = outer_model.loc[:, "redundancy"].loc[config.mvs(lv)].mean()
            if config.mode(lv) == Mode.A:
                ave_numerator = communality.sum()
                ave_denominator = ave_numerator + (1 - communality).sum() # = N
                ave.loc[lv] = ave_numerator / ave_denominator
                
                cr_numerator  = (loading.sum())**2
                cr_denominator = cr_numerator + (1-loading**2).sum()
                cr.loc[lv] = cr_numerator / cr_denominator
            if len(config.mvs(lv)) > 1:
                num_mvs_in_lv.append(len(config.mvs(lv)))
                communality_aux.append(block_communality.loc[lv])
        self.__summary = pd.concat([lv_type_text, r_squared, r_squared_adj, block_communality, mean_redundancy, ave, cr], axis=1,
                                   sort=True)
       

Remove comments if needed

[jezhumble]

Thanks for the suggestion. A couple of things.

1. I believe CR can be calculated for both modes A and B, but only if there is more than one MV defined. Therefore I think the calculation belongs after the line if len(config.mvs(lv)) > 1:
2. We will need to add a test for this, and update existing tests which will currently fail when this code is added. Which test we update will depend on where you get your calculation from. If you use seminr for example, we'd want to add it to the seminr regression test - currently that doesn't check the inner summary, but this would be a good opportunity to do that.

If you'd like to take on that work I strongly suggest you create a pull request to do it. Otherwise I am happy to take it on, but it won't be immediately.

Thanks again!