Latent_Profile_Analysis.qmd

---
title: Latent Profile Analysis
format: 
    gfm:
        toc: true
warning: false
fig-align: center
fig-cap-location: bottom
fig-width: 9
fig-height: 9
---



```{r}
#| include: false

library (reticulate)
path <- Sys.which("python")
path <- gsub("\\", "//", path, fixed = TRUE)
use_python(path)


```



## Using R (mclust package)


```{r}

# install.packages("mclust")
library(mclust)
library(tidyverse)

dat <- USArrests
dat |> dim()
dat |> head()
dat_scale <- dat |> scale() |> as.data.frame()
lpa_Model <- dat_scale |>  
                Mclust(G = 1:5)

lpa_Model |> summary()

lpa_assignments <- lpa_Model$classification

# Summarize the latent profiles
dat |> 
    scale() |> 
        as.data.frame() |> 
            mutate(Profile = lpa_assignments) |> 
                group_by(Profile) |> 
                    summarize_all(mean)
plot(lpa_Model)
pred_Model_1 <- lpa_Model$classification 

```


***
***

## Using R (tidyLPA package)


```{r}

# install.packages("tidyLPA")
n_profiles <- 3 # Specify the number of profiles

# Perform the LPA
lpa_Model2 <- dat |> 
                scale() |> 
                    as.data.frame() |> 
                    dplyr :: select(Murder, Assault, UrbanPop, Rape) |> 
                        tidyLPA :: estimate_profiles(n_profiles, variances = "equal")

# Print the results
print(lpa_Model2)


```



***
***
***
***


## Using python (scikit-learn)


```{python}
# first in terminal: 
# pip install -U scikit-learn


# Import the necessary libraries
import pandas as pd
import numpy as np
from sklearn.mixture import GaussianMixture

# Load the USArrests data
df = r.dat_scale
df.head()


# Perform the LPA using GaussianMixture
gmm = GaussianMixture(n_components=3)
gmm.fit(df)

# Print the results
print('Converged:',gmm.converged_) # Check if the model has converged
print('Means of each component:', gmm.means_)
print('Covariances of each component:', gmm.covariances_)

# Predict the labels for the data samples in df_scaled using gmm model
python_labels = gmm.predict(df)
py_labels = np.array(python_labels) + 1
R_labels = r.pred_Model_1
pd.DataFrame({"R_Labels": list(map(lambda x: int(x), R_labels)), "Python_Labels": py_labels})
```



***
***


## compare results


```{r}
adjustedRandIndex(py$R_labels, py$python_labels)
```