I am trying to run a Latent Class Analysis (LCA) after estimating the WTP values for 4 attributes. The purpose of the LCA is to create groups of people with differing WTP values, and including other covariates related to pro-environmental behavior (measured on a Likert type scale). It is my understanding that LCA via lclogit2 can handle non-binary data. When I run LCA with my attributes, the LCA works perfectly. When I add any other covariates, it just states values of '0.00' between classes. I have also tried turning the Likert type scale data into binary, but produced the same result. Is there a way to mitigate this? Or, perhaps just to save the class membership from the attributes and run a post-comparison model to see how the covariates vary per LCA class?
Attributes of WTP model: tax, rec_some, rec_all, eh_high, wq_inc, ei_pos [choice is their chosen attribute at given level]
Covariates [coded to binary data]: ct_binary, v_binary, s_binary
lclogit2 choice, rand( tax rec_some rec_all eh_high wq_inc ei_pos ct_binary v_binary s_binary ) id(id) group(setid) nclasses(4) iter(8)
Iteration 0: log likelihood = -10455.088
Iteration 1: log likelihood = -10415.537
Iteration 2: log likelihood = -10270.384
Iteration 3: log likelihood = -9974.4944
Iteration 4: log likelihood = -9713.9183
Iteration 5: log likelihood = -9551.6619
Iteration 6: log likelihood = -9459.546
Iteration 7: log likelihood = -9409.2248
Iteration 8: log likelihood = -9378.7617
The maximum number of iterations has been reached.
Latent class model with 4 latent classes
--------------------------------------------------
Variable | Class1 Class2 Class3 Class4
-------------+------------------------------------
tax | -0.021 -0.009 0.006 -0.001
rec_some | -0.335 0.339 0.173 0.906
rec_all | 0.246 0.338 0.170 0.896
eh_high | -0.134 0.852 0.508 -0.555
wq_inc | 0.689 0.941 0.645 -0.027
ei_pos | 0.390 0.737 0.651 1.034
o.ct_binary | 0.000 0.000 0.000 0.000
o.v_binary | 0.000 0.000 0.000 0.000
o.s_binary | 0.000 0.000 0.000 0.000
-------------+------------------------------------
Class Share | 0.239 0.179 0.374 0.208
--------------------------------------------------
Note: Model estimated via EM algorithm
. matrix start= e(b)
Attributes of WTP model: tax, rec_some, rec_all, eh_high, wq_inc, ei_pos [choice is their chosen attribute at given level]
Covariates [coded to binary data]: ct_binary, v_binary, s_binary
lclogit2 choice, rand( tax rec_some rec_all eh_high wq_inc ei_pos ct_binary v_binary s_binary ) id(id) group(setid) nclasses(4) iter(8)
Iteration 0: log likelihood = -10455.088
Iteration 1: log likelihood = -10415.537
Iteration 2: log likelihood = -10270.384
Iteration 3: log likelihood = -9974.4944
Iteration 4: log likelihood = -9713.9183
Iteration 5: log likelihood = -9551.6619
Iteration 6: log likelihood = -9459.546
Iteration 7: log likelihood = -9409.2248
Iteration 8: log likelihood = -9378.7617
The maximum number of iterations has been reached.
Latent class model with 4 latent classes
--------------------------------------------------
Variable | Class1 Class2 Class3 Class4
-------------+------------------------------------
tax | -0.021 -0.009 0.006 -0.001
rec_some | -0.335 0.339 0.173 0.906
rec_all | 0.246 0.338 0.170 0.896
eh_high | -0.134 0.852 0.508 -0.555
wq_inc | 0.689 0.941 0.645 -0.027
ei_pos | 0.390 0.737 0.651 1.034
o.ct_binary | 0.000 0.000 0.000 0.000
o.v_binary | 0.000 0.000 0.000 0.000
o.s_binary | 0.000 0.000 0.000 0.000
-------------+------------------------------------
Class Share | 0.239 0.179 0.374 0.208
--------------------------------------------------
Note: Model estimated via EM algorithm
. matrix start= e(b)
