Latent Profile Analysis of Duration Data
Functions
The functions to be used for the SDID and SC Estimates were created.
df <- read.csv("All Modules Duration Merged.csv")Models used
# Model 1: Equal variances, covariances = 0
# Model 2: Unequal variances, covariances = 0
# Model 3: Equal variances, equal covariances
# Model 6: Unequal variances, unequal covariancesEstimating the best number of classes using Model 1
From the estimation below, the result shows the Model 1 with 6 Classes is the best fit, but as observed in the result there is a warning from class 4 to class 6.
df1 <- df %>%
select(learnhpv_duration, learnvac_duration, getvac_duration, getans_duration) %>%
single_imputation() %>%
scale() %>%
estimate_profiles(1:6)
df1## tidyLPA analysis using mclust:
##
## Model Classes AIC BIC Entropy prob_min prob_max n_min n_max BLRT_p
## 1 1 8854.82 8892.09 1.00 1.00 1.00 1.00 1.00
## 1 2 8052.63 8113.18 0.94 0.91 0.99 0.10 0.90 0.01
## 1 3 7481.68 7565.53 0.93 0.91 0.99 0.02 0.79 0.01
## 1 4 7491.68 7598.82 0.61 0.00 0.99 0.00 0.79 0.63
## 1 5 7501.69 7632.11 0.41 0.00 0.99 0.00 0.77 0.84
## 1 6 7355.62 7509.33 0.49 0.00 1.00 0.00 0.55 0.01compare_solutions(df1, statistics = c("AIC", "BIC"))## Compare tidyLPA solutions:
##
## Model Classes AIC BIC Warnings
## 1 1 8854.822 8892.086
## 1 2 8052.627 8113.181
## 1 3 7481.684 7565.528
## 1 4 7491.683 7598.818 Warning
## 1 5 7501.686 7632.111 Warning
## 1 6 7355.617 7509.332 Warning
##
## Best model according to AIC is Model 1 with 6 classes.
## Best model according to BIC is Model 1 with 6 classes.
##
## An analytic hierarchy process, based on the fit indices AIC, AWE, BIC, CLC, and KIC (Akogul & Erisoglu, 2017), suggests the best solution is Model 1 with 6 classes.# More Statistics Check
compare_solutions(df1, statistics = c("AIC", "BIC", "AWE", "CLC", "KIC"))## Compare tidyLPA solutions:
##
## Model Classes AIC BIC AWE CLC KIC Warnings
## 1 1 8854.822 8892.086 8967.351 8840.822 8865.822
## 1 2 8052.627 8113.181 8236.860 8028.503 8068.627
## 1 3 7481.684 7565.528 7737.522 7447.534 7502.684
## 1 4 7491.683 7598.818 7819.726 7446.909 7517.683 Warning
## 1 5 7501.686 7632.111 7901.710 7446.511 7532.686 Warning
## 1 6 7355.617 7509.332 7827.058 7290.605 7391.617 Warning
##
## Best model according to AIC is Model 1 with 6 classes.
## Best model according to BIC is Model 1 with 6 classes.
## Best model according to AWE is Model 1 with 3 classes.
## Best model according to CLC is Model 1 with 6 classes.
## Best model according to KIC is Model 1 with 6 classes.
##
## An analytic hierarchy process, based on the fit indices AIC, AWE, BIC, CLC, and KIC (Akogul & Erisoglu, 2017), suggests the best solution is Model 1 with 6 classes.plot_profiles(df1)
plot_density(df1)
Estimating more classes using Model 1
The warning messages continues from class 4 upwards
df2 <- df %>%
select(learnhpv_duration, learnvac_duration, getvac_duration, getans_duration) %>%
single_imputation() %>%
scale() %>%
estimate_profiles(1:20)
df2## tidyLPA analysis using mclust:
##
## Model Classes AIC BIC Entropy prob_min prob_max n_min n_max BLRT_p
## 1 1 8854.82 8892.09 1.00 1.00 1.00 1.00 1.00
## 1 2 8052.63 8113.18 0.94 0.91 0.99 0.10 0.90 0.01
## 1 3 7481.68 7565.53 0.93 0.91 0.99 0.02 0.79 0.01
## 1 4 7491.68 7598.82 0.61 0.00 0.99 0.00 0.79 0.70
## 1 5 7501.69 7632.11 0.41 0.00 0.99 0.00 0.77 0.92
## 1 6 7355.62 7509.33 0.49 0.00 1.00 0.00 0.55 0.01
## 1 7 6921.76 7098.77 0.54 0.00 1.00 0.00 0.72 0.01
## 1 8 6931.76 7132.05 0.47 0.00 1.00 0.00 0.71 0.01
## 1 9 6600.12 6823.71 0.50 0.00 1.00 0.00 0.72 0.01
## 1 10 6479.90 6726.78 0.53 0.00 1.00 0.00 0.70 0.01
## 1 11 6272.10 6542.27 0.57 0.00 1.00 0.00 0.68 0.01
## 1 12 6231.08 6524.54 0.58 0.00 1.00 0.00 0.59 0.01
## 1 13 6128.28 6445.03 0.60 0.00 1.00 0.00 0.57 0.01
## 1 14 6138.28 6478.32 0.59 0.00 1.00 0.00 0.57 0.52
## 1 15 6148.29 6511.61 0.56 0.00 1.00 0.00 0.57 0.65
## 1 16 6286.35 6672.97 0.58 0.00 1.00 0.00 0.58 1.00
## 1 17 6498.85 6908.75 0.59 0.00 1.00 0.00 0.57 1.00
## 1 18 6508.71 6941.90 0.57 0.00 1.00 0.00 0.57 0.94
## 1 19 6518.87 6975.35 0.55 0.00 1.00 0.00 0.58 0.60
## 1 20 6528.88 7008.65 0.54 0.00 1.00 0.00 0.58 0.89compare_solutions(df2, statistics = c("AIC", "BIC"))## Compare tidyLPA solutions:
##
## Model Classes AIC BIC Warnings
## 1 1 8854.822 8892.086
## 1 2 8052.627 8113.181
## 1 3 7481.684 7565.528
## 1 4 7491.683 7598.818 Warning
## 1 5 7501.686 7632.111 Warning
## 1 6 7355.617 7509.332 Warning
## 1 7 6921.761 7098.766 Warning
## 1 8 6931.759 7132.053 Warning
## 1 9 6600.122 6823.707 Warning
## 1 10 6479.902 6726.776 Warning
## 1 11 6272.104 6542.269 Warning
## 1 12 6231.081 6524.536 Warning
## 1 13 6128.283 6445.028 Warning
## 1 14 6138.283 6478.318 Warning
## 1 15 6148.286 6511.611 Warning
## 1 16 6286.351 6672.966 Warning
## 1 17 6498.847 6908.752 Warning
## 1 18 6508.709 6941.904 Warning
## 1 19 6518.866 6975.351 Warning
## 1 20 6528.879 7008.654 Warning
##
## Best model according to AIC is Model 1 with 13 classes.
## Best model according to BIC is Model 1 with 13 classes.
##
## An analytic hierarchy process, based on the fit indices AIC, AWE, BIC, CLC, and KIC (Akogul & Erisoglu, 2017), suggests the best solution is Model 1 with 13 classes.# More Statistics Check
compare_solutions(df2, statistics = c("AIC", "BIC", "AWE", "CLC", "KIC"))## Compare tidyLPA solutions:
##
## Model Classes AIC BIC AWE CLC KIC Warnings
## 1 1 8854.822 8892.086 8967.351 8840.822 8865.822
## 1 2 8052.627 8113.181 8236.860 8028.503 8068.627
## 1 3 7481.684 7565.528 7737.522 7447.534 7502.684
## 1 4 7491.683 7598.818 7819.726 7446.909 7517.683 Warning
## 1 5 7501.686 7632.111 7901.710 7446.511 7532.686 Warning
## 1 6 7355.617 7509.332 7827.058 7290.605 7391.617 Warning
## 1 7 6921.761 7098.766 7464.700 6846.831 6962.761 Warning
## 1 8 6931.759 7132.053 7546.411 6846.695 6977.759 Warning
## 1 9 6600.122 6823.707 7286.290 6505.123 6651.122 Warning
## 1 10 6479.902 6726.776 7237.587 6374.965 6535.902 Warning
## 1 11 6272.104 6542.269 7101.287 6157.251 6333.104 Warning
## 1 12 6231.081 6524.536 7131.833 6106.239 6297.081 Warning
## 1 13 6128.283 6445.028 7100.574 5993.482 6199.283 Warning
## 1 14 6138.283 6478.318 7182.182 5993.454 6214.283 Warning
## 1 15 6148.286 6511.611 7263.807 5993.415 6229.286 Warning
## 1 16 6286.351 6672.966 7473.424 6121.507 6372.351 Warning
## 1 17 6498.847 6908.752 7757.467 6324.036 6589.847 Warning
## 1 18 6508.709 6941.904 7838.957 6323.851 6604.709 Warning
## 1 19 6518.866 6975.351 7920.734 6323.968 6619.866 Warning
## 1 20 6528.879 7008.654 8002.342 6323.966 6634.879 Warning
##
## Best model according to AIC is Model 1 with 13 classes.
## Best model according to BIC is Model 1 with 13 classes.
## Best model according to AWE is Model 1 with 13 classes.
## Best model according to CLC is Model 1 with 15 classes.
## Best model according to KIC is Model 1 with 13 classes.
##
## An analytic hierarchy process, based on the fit indices AIC, AWE, BIC, CLC, and KIC (Akogul & Erisoglu, 2017), suggests the best solution is Model 1 with 13 classes.plot_profiles(df2)
plot_density(df2)
Estimating classes without warning in Model 1
The estimation suggests that Model 1 with 3 classes is the best fit for our data and variables.
df3 <- df %>%
select(learnhpv_duration, learnvac_duration, getvac_duration, getans_duration) %>%
single_imputation() %>%
scale() %>%
estimate_profiles(1:3)
df3## tidyLPA analysis using mclust:
##
## Model Classes AIC BIC Entropy prob_min prob_max n_min n_max BLRT_p
## 1 1 8854.82 8892.09 1.00 1.00 1.00 1.00 1.00
## 1 2 8052.63 8113.18 0.94 0.91 0.99 0.10 0.90 0.01
## 1 3 7481.68 7565.53 0.93 0.91 0.99 0.02 0.79 0.01compare_solutions(df3, statistics = c("AIC", "BIC"))## Compare tidyLPA solutions:
##
## Model Classes AIC BIC
## 1 1 8854.822 8892.086
## 1 2 8052.627 8113.181
## 1 3 7481.684 7565.528
##
## Best model according to AIC is Model 1 with 3 classes.
## Best model according to BIC is Model 1 with 3 classes.
##
## An analytic hierarchy process, based on the fit indices AIC, AWE, BIC, CLC, and KIC (Akogul & Erisoglu, 2017), suggests the best solution is Model 1 with 3 classes.# More Statistics Check
compare_solutions(df3, statistics = c("AIC", "BIC", "AWE", "CLC", "KIC"))## Compare tidyLPA solutions:
##
## Model Classes AIC BIC AWE CLC KIC
## 1 1 8854.822 8892.086 8967.351 8840.822 8865.822
## 1 2 8052.627 8113.181 8236.860 8028.503 8068.627
## 1 3 7481.684 7565.528 7737.522 7447.534 7502.684
##
## Best model according to AIC is Model 1 with 3 classes.
## Best model according to BIC is Model 1 with 3 classes.
## Best model according to AWE is Model 1 with 3 classes.
## Best model according to CLC is Model 1 with 3 classes.
## Best model according to KIC is Model 1 with 3 classes.
##
## An analytic hierarchy process, based on the fit indices AIC, AWE, BIC, CLC, and KIC (Akogul & Erisoglu, 2017), suggests the best solution is Model 1 with 3 classes.plot_profiles(df3)
plot_density(df3)
Estimating using multiple models and classes
The estimation below suggests that Model 2 and 6 with 6 classes are the best fit.
df4 <- df %>%
select(learnhpv_duration, learnvac_duration, getvac_duration, getans_duration) %>%
single_imputation() %>%
scale() %>%
estimate_profiles(1:6, models = c(1, 2, 3, 6))## The 'variances'/'covariances' arguments were ignored in favor of the 'models' argument.df4## tidyLPA analysis using mclust:
##
## Model Classes AIC BIC Entropy prob_min prob_max n_min n_max BLRT_p
## 1 1 8854.82 8892.09 1.00 1.00 1.00 1.00 1.00
## 1 2 8052.63 8113.18 0.94 0.91 0.99 0.10 0.90 0.01
## 1 3 7481.68 7565.53 0.93 0.91 0.99 0.02 0.79 0.01
## 1 4 7491.68 7598.82 0.61 0.00 0.99 0.00 0.79 0.64
## 1 5 7501.69 7632.11 0.41 0.00 0.99 0.00 0.77 0.97
## 1 6 7355.62 7509.33 0.49 0.00 1.00 0.00 0.55 0.01
## 2 1 8854.82 8892.09 1.00 1.00 1.00 1.00 1.00
## 2 2 5842.68 5921.86 0.91 0.97 0.98 0.40 0.60 0.01
## 2 3 4963.72 5084.83 0.88 0.92 0.97 0.21 0.45 0.01
## 2 4 4570.03 4733.06 0.86 0.89 0.96 0.10 0.36 0.01
## 2 5 4404.89 4609.84 0.83 0.87 0.95 0.08 0.31 0.01
## 2 6 4309.37 4556.25 0.83 0.74 0.95 0.05 0.28 0.01
## 3 1 8052.09 8117.30 1.00 1.00 1.00 1.00 1.00
## 3 2 7449.89 7538.39 0.99 0.97 1.00 0.03 0.97 0.01
## 3 3 7459.89 7571.68 0.49 0.00 1.00 0.00 0.97 0.01
## 3 4 7260.24 7395.33 0.60 0.00 1.00 0.00 0.94 0.01
## 3 5 7270.24 7428.62 0.36 0.00 1.00 0.00 0.94 0.01
## 3 6 6971.03 7152.69 0.43 0.00 0.99 0.00 0.92 0.01
## 6 1 8052.09 8117.30 1.00 1.00 1.00 1.00 1.00
## 6 2 5289.68 5424.77 0.89 0.95 0.98 0.32 0.68 0.01
## 6 3 4742.37 4947.32 0.85 0.90 0.95 0.18 0.53 0.01
## 6 4 4462.24 4737.06 0.82 0.88 0.96 0.12 0.39 0.01
## 6 5 4286.92 4631.61 0.82 0.84 0.98 0.12 0.31 0.01
## 6 6 4186.14 4600.70 0.82 0.82 0.98 0.08 0.30 0.01compare_solutions(df4, statistics = c("AIC", "BIC"))## Compare tidyLPA solutions:
##
## Model Classes AIC BIC Warnings
## 1 1 8854.822 8892.086
## 1 2 8052.627 8113.181
## 1 3 7481.684 7565.528
## 1 4 7491.683 7598.818 Warning
## 1 5 7501.686 7632.111 Warning
## 1 6 7355.617 7509.332 Warning
## 2 1 8854.822 8892.086
## 2 2 5842.676 5921.862
## 2 3 4963.725 5084.833
## 2 4 4570.033 4733.064
## 2 5 4404.888 4609.840
## 2 6 4309.374 4556.249
## 3 1 8052.088 8117.300
## 3 2 7449.892 7538.394
## 3 3 7459.888 7571.680 Warning
## 3 4 7260.244 7395.326 Warning
## 3 5 7270.243 7428.616 Warning
## 3 6 6971.030 7152.693 Warning
## 6 1 8052.088 8117.300
## 6 2 5289.685 5424.767
## 6 3 4742.371 4947.324
## 6 4 4462.238 4737.061
## 6 5 4286.918 4631.611
## 6 6 4186.137 4600.700
##
## Best model according to AIC is Model 6 with 6 classes.
## Best model according to BIC is Model 2 with 6 classes.
##
## An analytic hierarchy process, based on the fit indices AIC, AWE, BIC, CLC, and KIC (Akogul & Erisoglu, 2017), suggests the best solution is Model 6 with 6 classes.# More Statistics Check
compare_solutions(df4, statistics = c("AIC", "BIC", "AWE", "CLC", "KIC"))## Compare tidyLPA solutions:
##
## Model Classes AIC BIC AWE CLC KIC Warnings
## 1 1 8854.822 8892.086 8967.351 8840.822 8865.822
## 1 2 8052.627 8113.181 8236.860 8028.503 8068.627
## 1 3 7481.684 7565.528 7737.522 7447.534 7502.684
## 1 4 7491.683 7598.818 7819.726 7446.909 7517.683 Warning
## 1 5 7501.686 7632.111 7901.710 7446.511 7532.686 Warning
## 1 6 7355.617 7509.332 7827.058 7290.605 7391.617 Warning
## 2 1 8854.822 8892.086 8967.351 8840.822 8865.822
## 2 2 5842.676 5921.862 6084.237 5810.487 5862.676
## 2 3 4963.725 5084.833 5334.184 4913.481 4992.725
## 2 4 4570.033 4733.064 5069.368 4501.760 4608.033
## 2 5 4404.888 4609.840 5033.135 4318.546 4451.888
## 2 6 4309.374 4556.249 5066.466 4205.031 4365.374
## 3 1 8052.088 8117.300 8250.512 8026.088 8069.088
## 3 2 7449.892 7538.394 7719.911 7413.878 7471.892
## 3 3 7459.888 7571.680 7802.494 7412.867 7486.888 Warning
## 3 4 7260.244 7395.326 7674.211 7203.441 7292.244 Warning
## 3 5 7270.243 7428.616 7756.276 7202.955 7307.243 Warning
## 3 6 6971.030 7152.693 7528.491 6893.895 7013.030 Warning
## 6 1 8052.088 8117.300 8250.512 8026.088 8069.088
## 6 2 5289.685 5424.767 5703.077 5233.458 5321.685
## 6 3 4742.371 4947.324 5370.580 4656.067 4789.371
## 6 4 4462.238 4737.061 5305.240 4345.881 4524.238
## 6 5 4286.918 4631.611 5344.661 4140.561 4363.918
## 6 6 4186.137 4600.700 5458.623 4009.776 4278.137
##
## Best model according to AIC is Model 6 with 6 classes.
## Best model according to BIC is Model 2 with 6 classes.
## Best model according to AWE is Model 2 with 5 classes.
## Best model according to CLC is Model 6 with 6 classes.
## Best model according to KIC is Model 6 with 6 classes.
##
## An analytic hierarchy process, based on the fit indices AIC, AWE, BIC, CLC, and KIC (Akogul & Erisoglu, 2017), suggests the best solution is Model 6 with 6 classes.plot_profiles(df4)
plot_density(df4)
Conclusion
The best Models and class for our data based on AIC & BIC statistics would be: #### 1. Model 1 with 3 Classes #### 2. Model 2 with 6 Classes #### 3. Model 6 with 6 Classes