I have a sample of 587 and am trying to do LPA based on protective factors (9 total). When I first ran the LPA, it seemed that 2 classes was the best fit with two groups. Per my advisor, I took out some of the protective factors to see if the profiles differed. Oddly, the numbers (N) in each class do not vary based on these changes. Could something be wrong with my data?
Secondly, is it ever acceptable to pick a class solution and then drop one of the classes due to small N? for example, one of my solutions is a 4 class solution that seems to fit well, but one group only has 6 members.