Dear all,
I am trying to build an index of e-transparency of organizations thanks to binary data (presence or absence of pieces of information on the website of the organization). I have 113 items which represent a piece of information each. Thanks to the literature, I have spotted 6 dimensions composing e-transparency. I would like to structure my data by performing a factor analysis, an exploratory one as I do not have assumptions on the model.
My objectives are to select items that are relevant to explain e-transparency (drop the irrelevant ones), group items according the dimension of transparency they are related to and weight each items according their importance within the dimension as well as weight the dimensions regarding e-transparency (it is a two levels latent factor model).
As my items are binary, I perfom an EFA on the tetrachoric correlation matrix. I therefore perform the following code:
However, I obtain the following error message: "tetra invalid; correlation outside [-1,1] found"
My first question: How can I solve this problem? When I check the values of my tetrachoric correlation matrix, it seems that there is no value outside [-1,1].
My second question: Some items are not relevant for some organizations. For instance, I check whether organizations give, or give not, the number of employees they have. If the organizations have no employee but only voluntary people, the organization cannot give its number of employee. The information doesn't appear on the website but on the aspect of transparency, it should not be penalized. Up to now, instead of 1 or 0, the value attributed to those situations is NA. How should I treat those values?
Thank you all of you for your help!
I am trying to build an index of e-transparency of organizations thanks to binary data (presence or absence of pieces of information on the website of the organization). I have 113 items which represent a piece of information each. Thanks to the literature, I have spotted 6 dimensions composing e-transparency. I would like to structure my data by performing a factor analysis, an exploratory one as I do not have assumptions on the model.
My objectives are to select items that are relevant to explain e-transparency (drop the irrelevant ones), group items according the dimension of transparency they are related to and weight each items according their importance within the dimension as well as weight the dimensions regarding e-transparency (it is a two levels latent factor model).
As my items are binary, I perfom an EFA on the tetrachoric correlation matrix. I therefore perform the following code:
Code:
tetrachoric item*, posdef matrix tetra = r(Rho) sca nobs = r(N) factormat tetra, n(`=nobs')
My first question: How can I solve this problem? When I check the values of my tetrachoric correlation matrix, it seems that there is no value outside [-1,1].
My second question: Some items are not relevant for some organizations. For instance, I check whether organizations give, or give not, the number of employees they have. If the organizations have no employee but only voluntary people, the organization cannot give its number of employee. The information doesn't appear on the website but on the aspect of transparency, it should not be penalized. Up to now, instead of 1 or 0, the value attributed to those situations is NA. How should I treat those values?
Thank you all of you for your help!
Comment