Classification differences between mclust and TidyLPA

[Odir Rodríguez Villagra]

Hello, 

First of all, thank you for this great package. I am a beginner on this technique (latent profile analysis) and I would like to ask you why the classification of observations between tidyLPA and mclust is not the same?  For example:

TidyLPA 

#Imputation

pisa_df<-pisaUSA15[1:100, ] %>%

  select(broad_interest, enjoyment, self_efficacy) %>%

  single_imputation() 

#Fitting

pisa_fit<-pisa_df %>%

  select(broad_interest, enjoyment, self_efficacy) %>%

  estimate_profiles(3,

                  models = c(1))

pisa_df_fitted<-get_data(pisa_fit)

pisa_df_fitted$Class<- as.factor(pisa_df_fitted$Class)

summary(pisa_df_fitted$Class)

> summary(pisa_df_fitted$Class)

 1  2  3 

29  3 68 

mclust (version 5.4.6)

pisa_mclust_fit <- Mclust(pisa, G = 3, modelName = "EEI")

pisa_mclust_fit$G

pisa_mclust_fit$modelName

table(summary(pisa_mclust_fit)$classification)

> table(summary(pisa_mclust_fit)$classification)

 1  2  3 

32  3 65 

In this example there are not so many differences between memberships but I have found important differences in other models. Also I found that the CPROBXs are different between packages.

Do you have any advice or recommendation to understand these differences?

I appreciate any help, sorry if this is a trivial question.

Best,