1 Pendahuluan

1.1 Latar Belakang

Media sosial telah menjadi bagian integral dari kehidupan modern, dengan lebih dari 5 miliar pengguna aktif di seluruh dunia. Namun di balik konektivitas yang ditawarkan, berbagai penelitian menunjukkan bahwa intensitas penggunaan media sosial berkorelasi dengan meningkatnya gejala kecemasan (anxiety), penurunan harga diri (self-esteem), dan menurunnya kesejahteraan psikologis (wellbeing) penggunanya.

Fenomena ini erat kaitannya dengan Social Comparison Theory (Festinger, 1954), yang menjelaskan kecenderungan individu untuk membandingkan dirinya dengan orang lain—sebuah perilaku yang diperkuat secara masif oleh fitur-fitur media sosial seperti likes, followers, dan highlight reels. Perbandingan sosial yang berlebihan ini dapat memperburuk persepsi diri dan memicu gangguan kesehatan mental.

Untuk memahami hubungan kausal antar variabel laten yang tidak dapat diukur secara langsung ini, diperlukan pendekatan analisis yang komprehensif. Structural Equation Modeling (SEM) merupakan teknik analisis multivariat yang menggabungkan analisis faktor konfirmatori (CFA) dengan analisis jalur (path analysis), sehingga memungkinkan pengujian hubungan kausal antar konstruk laten secara simultan dalam satu model berbasis teori.

1.2 Tujuan Penelitian

Mengukur pengaruh Social Media Use (SMU) terhadap Anxiety (ANX)
Mengukur pengaruh Social Media Use (SMU) terhadap Self-Esteem (SE)
Mengukur pengaruh Social Media Use (SMU), Anxiety (ANX), dan Self-Esteem (SE) terhadap Wellbeing (WB)
Memvalidasi model pengukuran (outer model) menggunakan Confirmatory Factor Analysis (CFA)
Menguji kesesuaian model struktural (inner model) secara keseluruhan

1.3 Hipotesis Penelitian

Kode	Jalur	Hipotesis
H1	SMU → ANX	Social Media Use berpengaruh positif terhadap Anxiety
H2	SMU → SE	Social Media Use berpengaruh negatif terhadap Self-Esteem
H3	SMU → WB	Social Media Use berpengaruh negatif terhadap Wellbeing
H4	ANX → WB	Anxiety berpengaruh negatif terhadap Wellbeing
H5	SE → WB	Self-Esteem berpengaruh positif terhadap Wellbeing

2 Dataset

2.1 Sumber & Deskripsi Data

Dataset yang digunakan adalah “Social Media and Mental Health” yang tersedia secara publik di Kaggle. Dataset ini dikumpulkan melalui survei untuk menginvestigasi korelasi antara intensitas penggunaan media sosial dengan kondisi kesehatan mental responden.

Sumber: Kaggle — souvikahmed071/social-media-and-mental-health
Jumlah observasi: 481 responden
Skala pengukuran: Skala Likert 1–5

2.2 Variabel & Konstruk Laten

Konstruk Laten	Kode	Indikator	Keterangan
Social Media Use	sm1	Q9	Penggunaan medsos tanpa tujuan
	sm2	Q10	Distraksi medsos saat sedang sibuk
	sm3	Q11	Gelisah jika tidak menggunakan medsos
	sm4	Q8	Rata-rata waktu harian di medsos (1–6)
Anxiety	anx1	Q12	Mudah terdistraksi
	anx2	Q13	Terganggu oleh kekhawatiran
	anx3	Q14	Sulit berkonsentrasi
Self-Esteem	se1	Q15	Membandingkan diri dengan orang sukses
	se2	Q16	Perasaan saat melakukan perbandingan
	se3	Q17	Mencari validasi dari fitur medsos
Wellbeing	wb1	Q18	Frekuensi merasa depresi/sedih
	wb2	Q19	Fluktuasi minat aktivitas harian
	wb3	Q20	Gangguan tidur

Semua konstruk menggunakan bentuk hubungan reflektif (konstruk → indikator), karena indikator merupakan cerminan dari konstruk latennya.

2.3 Model Konseptual

Social Media Use (SMU) ──→ Anxiety (ANX)
Social Media Use (SMU) ──→ Self-Esteem (SE)
Social Media Use (SMU) ──→ Wellbeing (WB)
           Anxiety (ANX) ──→ Wellbeing (WB)
        Self-Esteem (SE) ──→ Wellbeing (WB)

3 Persiapan

3.1 Load Library

# install.packages(c("lavaan","semPlot","psych","car",
#                    "corrplot","tidyverse","kableExtra"))

library(lavaan)
library(semTools)   # reliability() untuk objek lavaan (CR, AVE, omega)
library(semPlot)
library(psych)
library(car)
library(corrplot)
library(tidyverse)
library(kableExtra)

3.2 Import & Preprocessing Data

raw <- read.csv("smmh.csv", stringsAsFactors = FALSE)

# Rename kolom
colnames(raw) <- c(
  "timestamp", "age", "gender", "rel_status", "occupation",
  "org_type", "use_sm", "platforms", "daily_time",
  "Q9_purposeless", "Q10_distracted_busy", "Q11_restless",
  "Q12_easily_distracted", "Q13_bothered_worries", "Q14_diff_concentrate",
  "Q15_compare_others", "Q16_feel_comparison", "Q17_seek_validation",
  "Q18_feel_depressed", "Q19_interest_fluctuate", "Q20_sleep_issues"
)

# Encode daily_time → numerik ordinal 1–6
time_map <- c(
  "Less than an Hour"     = 1,
  "Between 1 and 2 hours" = 2,
  "Between 2 and 3 hours" = 3,
  "Between 3 and 4 hours" = 4,
  "Between 4 and 5 hours" = 5,
  "More than 5 hours"     = 6
)
raw$sm_time <- as.numeric(time_map[raw$daily_time])

# Filter pengguna medsos aktif & pilih indikator
df <- raw %>%
  filter(use_sm == "Yes") %>%
  select(
    sm1  = Q9_purposeless,
    sm2  = Q10_distracted_busy,
    sm3  = Q11_restless,
    sm4  = sm_time,
    anx1 = Q12_easily_distracted,
    anx2 = Q13_bothered_worries,
    anx3 = Q14_diff_concentrate,
    se1  = Q15_compare_others,
    se2  = Q16_feel_comparison,
    se3  = Q17_seek_validation,
    wb1  = Q18_feel_depressed,
    wb2  = Q19_interest_fluctuate,
    wb3  = Q20_sleep_issues
  ) %>%
  na.omit()

cat("Observasi setelah cleaning :", nrow(df), "\n")

## Observasi setelah cleaning : 478

cat("Jumlah variabel indikator   :", ncol(df), "\n")

## Jumlah variabel indikator   : 13

head(df, 5)

4 Uji Asumsi

4.1 1. Normalitas Multivariat (Mardia Test)

Salah satu asumsi CB-SEM adalah normalitas multivariat. Uji Mardia mengevaluasi skewness dan kurtosis secara multivariat. Pelanggaran asumsi ini diatasi dengan estimator MLR (Robust Maximum Likelihood dengan koreksi Satorra-Bentler).

mardia_result <- mardia(df, na.rm = TRUE)

data.frame(
  Test      = c("Mardia Skewness", "Mardia Kurtosis"),
  Statistic = c(round(mardia_result$b1p, 4), round(mardia_result$b2p, 4)),
  p_value   = c(round(mardia_result$p.skew, 4), round(mardia_result$p.kurt, 4)),
  Result    = c(
    ifelse(mardia_result$p.skew < 0.05, "Tidak Normal", "Normal"),
    ifelse(mardia_result$p.kurt < 0.05, "Tidak Normal", "Normal")
  )
) %>%
  kable(caption = "Uji Normalitas Multivariat — Mardia Test") %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"),
                full_width = FALSE) %>%
  column_spec(4, bold = TRUE,
              color = ifelse(
                c(mardia_result$p.skew, mardia_result$p.kurt) < 0.05,
                "red", "darkgreen"
              ))

Uji Normalitas Multivariat — Mardia Test
Test	Statistic	p_value	Result
Mardia Skewness	8.7619	0	Tidak Normal
Mardia Kurtosis	203.5982	0	Tidak Normal

Interpretasi: Jika p-value < 0.05 pada skewness atau kurtosis, data tidak berdistribusi normal multivariat. Estimator MLR digunakan sebagai solusi karena robust terhadap pelanggaran normalitas.

4.2 2. Kecukupan Data (KMO Test)

Nilai Kaiser-Meyer-Olkin (KMO) / Overall MSA > 0.5 menunjukkan korelasi antar indikator cukup kuat untuk analisis faktor.

r_matrix  <- cor(df)
kmo_result <- KMO(r_matrix)

cat("Overall MSA (KMO):", round(kmo_result$MSA, 4), "\n\n")

## Overall MSA (KMO): 0.8887

cat("MSA per indikator:\n")

## MSA per indikator:

print(round(kmo_result$MSAi, 3))

##   sm1   sm2   sm3   sm4  anx1  anx2  anx3   se1   se2   se3   wb1   wb2   wb3 
## 0.869 0.889 0.914 0.907 0.872 0.893 0.899 0.860 0.537 0.797 0.895 0.935 0.916

kmo_val <- round(kmo_result$MSA, 3)
kmo_ket <- case_when(
  kmo_val >= 0.90 ~ "Marvelous (sangat baik sekali)",
  kmo_val >= 0.80 ~ "Meritorious (sangat baik)",
  kmo_val >= 0.70 ~ "Middling (baik)",
  kmo_val >= 0.60 ~ "Mediocre (cukup)",
  kmo_val >= 0.50 ~ "Miserable (lemah tapi masih diterima)",
  TRUE            ~ "Unacceptable (tidak layak)"
)
cat(sprintf("KMO = %.3f → %s\n", kmo_val, kmo_ket))

## KMO = 0.889 → Meritorious (sangat baik)

4.3 3. Non-Multikolinieritas (VIF)

VIF < 3.3 menunjukkan tidak ada multikolinieritas antar indikator yang mengancam estimasi.

model_vif <- lm(wb1 ~ sm1 + sm2 + sm3 + sm4 +
                      anx1 + anx2 + anx3 +
                      se1 + se2 + se3 +
                      wb2 + wb3, data = df)

vif_values <- vif(model_vif)

data.frame(
  Indikator = names(vif_values),
  VIF       = round(vif_values, 3),
  Status    = ifelse(vif_values < 3.3, "Aman", "Bermasalah")
) %>%
  kable(caption = "Nilai VIF per Indikator (Threshold < 3.3)") %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"),
                full_width = FALSE) %>%
  column_spec(3, bold = TRUE,
              color = ifelse(vif_values < 3.3, "darkgreen", "red"))

Nilai VIF per Indikator (Threshold < 3.3)
	Indikator	VIF	Status
sm1	sm1	1.453	Aman
sm2	sm2	1.969	Aman
sm3	sm3	1.673	Aman
sm4	sm4	1.354	Aman
anx1	anx1	2.310	Aman
anx2	anx2	1.645	Aman
anx3	anx3	2.271	Aman
se1	se1	1.481	Aman
se2	se2	1.060	Aman
se3	se3	1.298	Aman
wb2	wb2	1.616	Aman
wb3	wb3	1.227	Aman

4.4 4. Matriks Korelasi Antar Indikator

corrplot(r_matrix,
         method      = "color",
         type        = "upper",
         order       = "hclust",
         tl.cex      = 0.8,
         tl.col      = "black",
         addCoef.col = "black",
         number.cex  = 0.58,
         title       = "Matriks Korelasi Antar Indikator",
         mar         = c(0, 0, 1.5, 0))

Indikator dalam satu konstruk yang sama seharusnya berkorelasi lebih tinggi satu sama lain dibandingkan dengan indikator dari konstruk yang berbeda.

5 Outer Model — Evaluasi per Konstruk

5.1 Fungsi Pembantu Evaluasi

eval_outer <- function(fit_obj, construct_name) {
  cat("--------------------------------------\n")
  cat(" Konstruk:", construct_name, "\n")
  cat("--------------------------------------\n\n")

  # Convergence guard — skip detailed output if model did not converge
  if (!lavaan::lavInspect(fit_obj, "converged")) {
    cat("[PERINGATAN] Model tidak konvergen. Periksa spesifikasi atau kualitas indikator.\n\n")
    return(invisible(NULL))
  }

  # Loading Factor
  cat(">> Loading Factor (Standardized)\n")
  sl <- standardizedsolution(fit_obj)
  lf <- sl[sl$op == "=~", c("lhs", "rhs", "est.std", "pvalue")]
  names(lf) <- c("Konstruk", "Indikator", "Loading Std", "p-value")
  lf$Status <- ifelse(abs(lf$`Loading Std`) >= 0.70, "Ideal",
               ifelse(abs(lf$`Loading Std`) >= 0.50, "Cukup", "Drop"))
  print(lf, row.names = FALSE)

  # Model Fit
  cat("\n>> Model Fit\n")
  fi <- fitMeasures(fit_obj, c("chisq", "df", "pvalue",
                                "cfi", "tli", "rmsea", "srmr"))
  fi_df <- data.frame(
    Indeks    = names(fi),
    Nilai     = round(fi, 4),
    Threshold = c("—", "—", "> 0.05", "≥ 0.90", "≥ 0.90", "< 0.08", "< 0.08"),
    Status    = c("—", "—",
                  ifelse(fi["pvalue"] > 0.05, "Lulus", "Gagal"),
                  ifelse(fi["cfi"]    >= 0.90, "Lulus", "Gagal"),
                  ifelse(fi["tli"]    >= 0.90, "Lulus", "Gagal"),
                  ifelse(fi["rmsea"]  <  0.08, "Lulus", "Gagal"),
                  ifelse(fi["srmr"]   <  0.08, "Lulus", "Gagal"))
  )
  print(fi_df, row.names = FALSE)

  # Reliabilitas & AVE (semTools::reliability menerima objek lavaan)
  rel <- semTools::reliability(fit_obj)
  cat("\n>> Reliabilitas & Validitas Konvergen\n")
  rel_df <- data.frame(
    Ukuran = c("Cronbach Alpha", "Omega / CR", "AVE"),
    Nilai  = round(c(rel["alpha", 1], rel["omega", 1], rel["avevar", 1]), 4),
    Threshold = c("≥ 0.70", "≥ 0.70", "≥ 0.50"),
    Status = c(
      ifelse(rel["alpha",  1] >= 0.70, "Lulus", "Gagal"),
      ifelse(rel["omega",  1] >= 0.70, "Lulus", "Gagal"),
      ifelse(rel["avevar", 1] >= 0.50, "Lulus", "Gagal")
    )
  )
  print(rel_df, row.names = FALSE)
  cat("\n")
}

5.2 CFA: Social Media Use (SMU)

model_smu <- 'SMU =~ sm1 + sm2 + sm3 + sm4'
fit_smu   <- cfa(model_smu, data = df, estimator = "MLR")

eval_outer(fit_smu, "Social Media Use")

## --------------------------------------
##  Konstruk: Social Media Use 
## --------------------------------------
## 
## >> Loading Factor (Standardized)
##  Konstruk Indikator Loading.Std p.value Status
##       SMU       sm1       0.608       0  Cukup
##       SMU       sm2       0.735       0  Ideal
##       SMU       sm3       0.664       0  Cukup
##       SMU       sm4       0.563       0  Cukup
## 
## >> Model Fit
##  Indeks   Nilai Threshold Status
##   chisq 10.2158         —      —
##      df  2.0000         —      —
##  pvalue  0.0060    > 0.05  Gagal
##     cfi  0.9790    ≥ 0.90  Lulus
##     tli  0.9371    ≥ 0.90  Lulus
##   rmsea  0.0927    < 0.08  Gagal
##    srmr  0.0267    < 0.08  Lulus

## 
## >> Reliabilitas & Validitas Konvergen
##          Ukuran  Nilai Threshold Status
##  Cronbach Alpha 0.7276    ≥ 0.70  Lulus
##      Omega / CR 0.7310    ≥ 0.70  Lulus
##             AVE 0.4088    ≥ 0.50  Gagal

semPaths(fit_smu, what = "std", layout = "tree",
         edge.label.cex = 0.9, sizeMan = 8, sizeLat = 10,
         fade = FALSE, title = FALSE, mar = c(3, 3, 3, 3))
title("CFA — Social Media Use", line = 2)

5.3 CFA: Anxiety (ANX)

model_anx <- 'ANX =~ anx1 + anx2 + anx3'
fit_anx   <- cfa(model_anx, data = df, estimator = "MLR")

eval_outer(fit_anx, "Anxiety")

## --------------------------------------
##  Konstruk: Anxiety 
## --------------------------------------
## 
## >> Loading Factor (Standardized)
##  Konstruk Indikator Loading.Std p.value Status
##       ANX      anx1       0.745       0  Ideal
##       ANX      anx2       0.611       0  Cukup
##       ANX      anx3       0.884       0  Ideal
## 
## >> Model Fit
##  Indeks Nilai Threshold Status
##   chisq     0         —      —
##      df     0         —      —
##  pvalue    NA    > 0.05   <NA>
##     cfi     1    ≥ 0.90  Lulus
##     tli     1    ≥ 0.90  Lulus
##   rmsea     0    < 0.08  Lulus
##    srmr     0    < 0.08  Lulus

## 
## >> Reliabilitas & Validitas Konvergen
##          Ukuran  Nilai Threshold Status
##  Cronbach Alpha 0.7855    ≥ 0.70  Lulus
##      Omega / CR 0.7990    ≥ 0.70  Lulus
##             AVE 0.5780    ≥ 0.50  Lulus

semPaths(fit_anx, what = "std", layout = "tree",
         edge.label.cex = 0.9, sizeMan = 8, sizeLat = 10,
         fade = FALSE, title = FALSE, mar = c(3, 3, 3, 3))
title("CFA — Anxiety", line = 2)

5.4 CFA: Self-Esteem (SE)

model_se <- 'SE =~ se1 + se2 + se3'
# bounds = TRUE mencegah Heywood case (residual variance negatif) akibat
# korelasi rendah se2 (Q16 mengukur sentimen, bukan frekuensi seperti se1/se3)
fit_se   <- cfa(model_se, data = df, estimator = "MLR", bounds = TRUE)

eval_outer(fit_se, "Self-Esteem")

## --------------------------------------
##  Konstruk: Self-Esteem 
## --------------------------------------
## 
## >> Loading Factor (Standardized)
##  Konstruk Indikator Loading.Std p.value Status
##        SE       se1       0.402   0.000   Drop
##        SE       se2       0.152   0.001   Drop
##        SE       se3       1.025   0.000  Ideal
## 
## >> Model Fit
##  Indeks  Nilai Threshold Status
##   chisq 3.1419         —      —
##      df 0.0000         —      —
##  pvalue     NA    > 0.05   <NA>
##     cfi 0.9688    ≥ 0.90  Lulus
##     tli 1.0000    ≥ 0.90  Lulus
##   rmsea 0.0000    < 0.08  Lulus
##    srmr 0.0301    < 0.08  Lulus

## 
## >> Reliabilitas & Validitas Konvergen
##          Ukuran  Nilai Threshold Status
##  Cronbach Alpha 0.4210    ≥ 0.70  Gagal
##      Omega / CR 0.6009    ≥ 0.70  Gagal
##             AVE 0.4259    ≥ 0.50  Gagal

semPaths(fit_se, what = "std", layout = "tree",
         edge.label.cex = 0.9, sizeMan = 8, sizeLat = 10,
         fade = FALSE, title = FALSE, mar = c(3, 3, 3, 3))
title("CFA — Self-Esteem", line = 2)

5.5 CFA: Wellbeing (WB)

model_wb <- 'WB =~ wb1 + wb2 + wb3'
fit_wb   <- cfa(model_wb, data = df, estimator = "MLR")

eval_outer(fit_wb, "Wellbeing")

## --------------------------------------
##  Konstruk: Wellbeing 
## --------------------------------------
## 
## >> Loading Factor (Standardized)
##  Konstruk Indikator Loading.Std p.value Status
##        WB       wb1       0.760       0  Ideal
##        WB       wb2       0.647       0  Cukup
##        WB       wb3       0.484       0   Drop
## 
## >> Model Fit
##  Indeks Nilai Threshold Status
##   chisq     0         —      —
##      df     0         —      —
##  pvalue    NA    > 0.05   <NA>
##     cfi     1    ≥ 0.90  Lulus
##     tli     1    ≥ 0.90  Lulus
##   rmsea     0    < 0.08  Lulus
##    srmr     0    < 0.08  Lulus

## 
## >> Reliabilitas & Validitas Konvergen
##          Ukuran  Nilai Threshold Status
##  Cronbach Alpha 0.6525    ≥ 0.70  Gagal
##      Omega / CR 0.6591    ≥ 0.70  Gagal
##             AVE 0.3966    ≥ 0.50  Gagal

semPaths(fit_wb, what = "std", layout = "tree",
         edge.label.cex = 0.9, sizeMan = 8, sizeLat = 10,
         fade = FALSE, title = FALSE, mar = c(3, 3, 3, 3))
title("CFA — Wellbeing", line = 2)

6 First Order CFA — Validitas Diskriminan

6.1 Spesifikasi & Estimasi Model

model_cfa1 <- '
  SMU =~ sm1 + sm2 + sm3 + sm4
  ANX =~ anx1 + anx2 + anx3
  SE  =~ se1 + se2 + se3
  WB  =~ wb1 + wb2 + wb3
'
fit_cfa1 <- cfa(model_cfa1, data = df, estimator = "MLR", bounds = TRUE)
summary(fit_cfa1, fit.measures = TRUE, standardized = TRUE)

## lavaan 0.6-21 ended normally after 46 iterations
## 
##   Estimator                                         ML
##   Optimization method                           NLMINB
##   Number of model parameters                        32
##   Row rank of the constraints matrix                32
## 
##   Number of observations                           478
## 
## Model Test User Model:
##                                               Standard      Scaled
##   Test Statistic                               222.216     205.961
##   Degrees of freedom                                59          59
##   P-value (Chi-square)                           0.000       0.000
##   Scaling correction factor                                  1.079
##     Yuan-Bentler correction (Mplus variant)                       
## 
## Model Test Baseline Model:
## 
##   Test statistic                              2036.771    1787.843
##   Degrees of freedom                                78          78
##   P-value                                        0.000       0.000
##   Scaling correction factor                                  1.139
## 
## User Model versus Baseline Model:
## 
##   Comparative Fit Index (CFI)                    0.917       0.914
##   Tucker-Lewis Index (TLI)                       0.890       0.886
##                                                                   
##   Robust Comparative Fit Index (CFI)                         0.919
##   Robust Tucker-Lewis Index (TLI)                            0.892
## 
## Loglikelihood and Information Criteria:
## 
##   Loglikelihood user model (H0)              -9450.234   -9450.234
##   Scaling correction factor                                  1.002
##       for the MLR correction                                      
##   Loglikelihood unrestricted model (H1)      -9339.126   -9339.126
##   Scaling correction factor                                  1.052
##       for the MLR correction                                      
##                                                                   
##   Akaike (AIC)                               18964.468   18964.468
##   Bayesian (BIC)                             19097.896   19097.896
##   Sample-size adjusted Bayesian (SABIC)      18996.332   18996.332
## 
## Root Mean Square Error of Approximation:
## 
##   RMSEA                                          0.076       0.072
##   90 Percent confidence interval - lower         0.066       0.062
##   90 Percent confidence interval - upper         0.087       0.083
##   P-value H_0: RMSEA <= 0.050                    0.000       0.000
##   P-value H_0: RMSEA >= 0.080                    0.283       0.112
##                                                                   
##   Robust RMSEA                                               0.075
##   90 Percent confidence interval - lower                     0.064
##   90 Percent confidence interval - upper                     0.086
##   P-value H_0: Robust RMSEA <= 0.050                         0.000
##   P-value H_0: Robust RMSEA >= 0.080                         0.239
## 
## Standardized Root Mean Square Residual:
## 
##   SRMR                                           0.048       0.048
## 
## Parameter Estimates:
## 
##   Standard errors                             Sandwich
##   Information bread                           Observed
##   Observed information based on                Hessian
## 
## Latent Variables:
##                    Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
##   SMU =~                                                                
##     sm1               1.000                               0.610    0.560
##     sm2               1.655    0.155   10.709    0.000    1.010    0.766
##     sm3               1.405    0.147    9.526    0.000    0.857    0.684
##     sm4               1.391    0.149    9.323    0.000    0.849    0.536
##   ANX =~                                                                
##     anx1              1.000                               0.910    0.778
##     anx2              0.922    0.076   12.159    0.000    0.839    0.657
##     anx3              1.202    0.064   18.830    0.000    1.094    0.816
##   SE =~                                                                 
##     se1               1.000                               1.062    0.757
##     se2               0.070    0.083    0.845    0.398    0.074    0.070
##     se3               0.636    0.104    6.116    0.000    0.675    0.542
##   WB =~                                                                 
##     wb1               1.000                               0.943    0.722
##     wb2               0.931    0.063   14.857    0.000    0.878    0.700
##     wb3               0.708    0.069   10.227    0.000    0.668    0.458
## 
## Covariances:
##                    Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
##   SMU ~~                                                                
##     ANX               0.461    0.055    8.330    0.000    0.830    0.830
##     SE                0.334    0.050    6.633    0.000    0.515    0.515
##     WB                0.418    0.056    7.474    0.000    0.727    0.727
##   ANX ~~                                                                
##     SE                0.545    0.070    7.744    0.000    0.564    0.564
##     WB                0.755    0.064   11.884    0.000    0.880    0.880
##   SE ~~                                                                 
##     WB                0.699    0.079    8.805    0.000    0.698    0.698
## 
## Variances:
##                    Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
##    .sm1               0.815    0.062   13.260    0.000    0.815    0.687
##    .sm2               0.718    0.081    8.866    0.000    0.718    0.413
##    .sm3               0.836    0.064   13.084    0.000    0.836    0.532
##    .sm4               1.786    0.112   15.934    0.000    1.786    0.712
##    .anx1              0.541    0.054   10.058    0.000    0.541    0.395
##    .anx2              0.927    0.086   10.820    0.000    0.927    0.569
##    .anx3              0.602    0.079    7.642    0.000    0.602    0.335
##    .se1               0.841    0.171    4.925    0.000    0.841    0.427
##    .se2               1.105    0.065   16.905    0.000    1.105    0.995
##    .se3               1.094    0.107   10.207    0.000    1.094    0.706
##    .wb1               0.818    0.075   10.883    0.000    0.818    0.479
##    .wb2               0.803    0.073   10.935    0.000    0.803    0.510
##    .wb3               1.676    0.099   17.013    0.000    1.676    0.790
##     SMU               0.372    0.068    5.494    0.000    1.000    1.000
##     ANX               0.828    0.083    9.945    0.000    1.000    1.000
##     SE                1.128    0.185    6.101    0.000    1.000    1.000
##     WB                0.889    0.097    9.145    0.000    1.000    1.000

6.2 Loading Factor Gabungan

std_sol <- standardizedsolution(fit_cfa1)
lf_all  <- std_sol[std_sol$op == "=~",
                   c("lhs", "rhs", "est.std", "se", "z", "pvalue")]
names(lf_all) <- c("Konstruk", "Indikator", "Loading Std", "SE", "z-value", "p-value")

lf_all %>%
  mutate(
    `Loading Std` = round(`Loading Std`, 3),
    SE            = round(SE, 3),
    `z-value`    = round(`z-value`, 3),
    `p-value`    = round(`p-value`, 4),
    Status       = case_when(
      abs(`Loading Std`) >= 0.70 ~ "Ideal",
      abs(`Loading Std`) >= 0.50 ~ "Cukup",
      TRUE                       ~ "Drop"
    )
  ) %>%
  kable(caption = "Loading Factor Standardized — First Order CFA") %>%
  kable_styling(bootstrap_options = c("striped", "hover")) %>%
  pack_rows("Social Media Use", 1, 4) %>%
  pack_rows("Anxiety", 5, 7) %>%
  pack_rows("Self-Esteem", 8, 10) %>%
  pack_rows("Wellbeing", 11, 13) %>%
  column_spec(7, bold = TRUE,
              color = ifelse(abs(lf_all$`Loading Std`) >= 0.70, "darkgreen",
                      ifelse(abs(lf_all$`Loading Std`) >= 0.50, "orange", "red")))

Loading Factor Standardized — First Order CFA
Konstruk	Indikator	Loading Std	SE	z-value	p-value	Status
Social Media Use
SMU	sm1	0.560	0.044	12.754	0.0000	Cukup
SMU	sm2	0.766	0.031	24.541	0.0000	Ideal
SMU	sm3	0.684	0.030	22.646	0.0000	Cukup
SMU	sm4	0.536	0.042	12.796	0.0000	Cukup
Anxiety
ANX	anx1	0.778	0.027	28.698	0.0000	Ideal
ANX	anx2	0.657	0.039	16.896	0.0000	Cukup
ANX	anx3	0.816	0.027	30.401	0.0000	Ideal
Self-Esteem
SE	se1	0.757	0.058	13.069	0.0000	Ideal
SE	se2	0.070	0.080	0.883	0.3771	Drop
SE	se3	0.542	0.060	9.082	0.0000	Cukup
Wellbeing
WB	wb1	0.722	0.031	22.930	0.0000	Ideal
WB	wb2	0.700	0.032	21.590	0.0000	Ideal
WB	wb3	0.458	0.044	10.342	0.0000	Drop

6.3 Model Fit — First Order CFA

fit_idx <- fitMeasures(fit_cfa1, c(
  "chisq", "df", "pvalue",
  "cfi", "tli",
  "rmsea", "rmsea.ci.lower", "rmsea.ci.upper", "srmr"
))

data.frame(
  Indeks    = names(fit_idx),
  Nilai     = round(fit_idx, 4),
  Threshold = c("—", "—", "> 0.05",
                "≥ 0.90", "≥ 0.90",
                "< 0.08", "—", "—", "< 0.08"),
  Status    = c("—", "—",
                ifelse(fit_idx["pvalue"] > 0.05, "Lulus", "Gagal"),
                ifelse(fit_idx["cfi"]    >= 0.90, "Lulus", "Gagal"),
                ifelse(fit_idx["tli"]    >= 0.90, "Lulus", "Gagal"),
                ifelse(fit_idx["rmsea"]  <  0.08, "Lulus", "Gagal"),
                "—", "—",
                ifelse(fit_idx["srmr"]   <  0.08, "Lulus", "Gagal"))
) %>%
  kable(caption = "Model Fit — First Order CFA") %>%
  kable_styling(bootstrap_options = c("striped", "hover"), full_width = FALSE) %>%
  column_spec(4, bold = TRUE,
              color = case_when(
                grepl("Lulus", c("—","—",
                                  ifelse(fit_idx["pvalue"] > 0.05, "Lulus","Gagal"),
                                  ifelse(fit_idx["cfi"]   >= 0.90, "Lulus","Gagal"),
                                  ifelse(fit_idx["tli"]   >= 0.90, "Lulus","Gagal"),
                                  ifelse(fit_idx["rmsea"] <  0.08, "Lulus","Gagal"),
                                  "—","—",
                                  ifelse(fit_idx["srmr"]  <  0.08, "Lulus","Gagal"))) ~ "darkgreen",
                grepl("Gagal", c("—","—",
                                  ifelse(fit_idx["pvalue"] > 0.05, "Lulus","Gagal"),
                                  ifelse(fit_idx["cfi"]   >= 0.90, "Lulus","Gagal"),
                                  ifelse(fit_idx["tli"]   >= 0.90, "Lulus","Gagal"),
                                  ifelse(fit_idx["rmsea"] <  0.08, "Lulus","Gagal"),
                                  "—","—",
                                  ifelse(fit_idx["srmr"]  <  0.08, "Lulus","Gagal"))) ~ "red",
                TRUE ~ "gray"
              ))

Model Fit — First Order CFA
	Indeks	Nilai	Threshold	Status
chisq	chisq	222.2163	—	—
df	df	59.0000	—	—
pvalue	pvalue	0.0000	> 0.05	Gagal
cfi	cfi	0.9167	≥ 0.90	Lulus
tli	tli	0.8898	≥ 0.90	Gagal
rmsea	rmsea	0.0761	< 0.08	Lulus
rmsea.ci.lower	rmsea.ci.lower	0.0656	—	—
rmsea.ci.upper	rmsea.ci.upper	0.0869	—	—
srmr	srmr	0.0478	< 0.08	Lulus

6.4 Reliabilitas & AVE — Semua Konstruk

rel_all <- semTools::reliability(fit_cfa1)

as.data.frame(t(rel_all)) %>%
  rownames_to_column("Konstruk") %>%
  select(Konstruk,
         `Cronbach Alpha` = alpha,
         `Omega (CR)`     = omega,
         AVE              = avevar) %>%
  mutate(
    across(where(is.numeric), ~ round(.x, 4)),
    `Status Alpha` = ifelse(`Cronbach Alpha` >= 0.70, "Lulus", "Gagal"),
    `Status CR`    = ifelse(`Omega (CR)`     >= 0.70, "Lulus", "Gagal"),
    `Status AVE`   = ifelse(AVE              >= 0.50, "Lulus", "Gagal")
  ) %>%
  kable(caption = "Reliabilitas Konstruk & Validitas Konvergen (AVE)") %>%
  kable_styling(bootstrap_options = c("striped", "hover"), full_width = FALSE)

Reliabilitas Konstruk & Validitas Konvergen (AVE)
Konstruk	Cronbach Alpha	Omega (CR)	AVE	Status Alpha	Status CR	Status AVE
SMU	0.7276	0.7270	0.4066	Lulus	Lulus	Gagal
ANX	0.7855	0.7961	0.5686	Lulus	Lulus	Lulus
SE	0.4210	0.5189	0.3432	Gagal	Gagal	Gagal
WB	0.6525	0.6525	0.3896	Gagal	Gagal	Gagal

6.5 Validitas Diskriminan

6.5.1 Fornell-Larcker Criterion

lv_cor   <- lavInspect(fit_cfa1, "cor.lv")
ave_vec  <- rel_all["avevar", ]
fl_mat   <- lv_cor
diag(fl_mat) <- sqrt(ave_vec)

round(fl_mat, 3) %>%
  kable(caption = "Fornell-Larcker Criterion — Diagonal = √AVE") %>%
  kable_styling(bootstrap_options = c("striped", "hover"), full_width = FALSE)

Fornell-Larcker Criterion — Diagonal = √AVE
	SMU	ANX	SE	WB
SMU	0.638	0.830	0.515	0.727
ANX	0.830	0.754	0.564	0.880
SE	0.515	0.564	0.586	0.698
WB	0.727	0.880	0.698	0.624

Kriteria: Nilai diagonal (√AVE) harus lebih besar dari semua nilai off-diagonal pada baris/kolom yang sama.

6.5.2 HTMT Ratio

htmt_result <- htmt(model_cfa1, data = df)

round(htmt_result, 3) %>%
  kable(caption = "HTMT Ratio — Threshold < 0.85 (ketat) atau < 0.90") %>%
  kable_styling(bootstrap_options = c("striped", "hover"), full_width = FALSE)

HTMT Ratio — Threshold < 0.85 (ketat) atau < 0.90
	SMU	ANX	SE	WB
SMU	1.000	0.786	0.746	0.718
ANX	0.786	1.000	0.513	0.909
SE	0.746	0.513	1.000	0.754
WB	0.718	0.909	0.754	1.000

Kriteria: Semua nilai HTMT < 0.85 menunjukkan validitas diskriminan terpenuhi.

6.6 Path Diagram — First Order CFA

semPaths(fit_cfa1,
         what           = "std",
         layout         = "tree2",
         rotation       = 2,
         edge.label.cex = 0.75,
         sizeMan        = 6,
         sizeLat        = 9,
         fade           = FALSE,
         title          = FALSE,
         mar            = c(2, 2, 2, 2))
title("First Order CFA — Semua Konstruk", line = 1.5)

7 Inner Model — Full SEM

7.1 Spesifikasi Model

model_sem <- '
  # ── OUTER MODEL ──────────────────────────────
  SMU =~ sm1 + sm2 + sm3 + sm4
  ANX =~ anx1 + anx2 + anx3
  SE  =~ se1 + se2 + se3
  WB  =~ wb1 + wb2 + wb3

  # ── INNER MODEL ──────────────────────────────
  ANX ~ SMU          # H1: SMU → ANX
  SE  ~ SMU          # H2: SMU → SE
  WB  ~ SMU + ANX + SE  # H3, H4, H5
'
fit_sem <- sem(model_sem, data = df, estimator = "MLR")

7.2 Ringkasan Hasil SEM

summary(fit_sem, fit.measures = TRUE, standardized = TRUE, rsquare = TRUE)

## lavaan 0.6-21 ended normally after 42 iterations
## 
##   Estimator                                         ML
##   Optimization method                           NLMINB
##   Number of model parameters                        31
## 
##   Number of observations                           478
## 
## Model Test User Model:
##                                               Standard      Scaled
##   Test Statistic                               229.683     212.454
##   Degrees of freedom                                60          60
##   P-value (Chi-square)                           0.000       0.000
##   Scaling correction factor                                  1.081
##     Yuan-Bentler correction (Mplus variant)                       
## 
## Model Test Baseline Model:
## 
##   Test statistic                              2036.771    1787.843
##   Degrees of freedom                                78          78
##   P-value                                        0.000       0.000
##   Scaling correction factor                                  1.139
## 
## User Model versus Baseline Model:
## 
##   Comparative Fit Index (CFI)                    0.913       0.911
##   Tucker-Lewis Index (TLI)                       0.887       0.884
##                                                                   
##   Robust Comparative Fit Index (CFI)                         0.915
##   Robust Tucker-Lewis Index (TLI)                            0.890
## 
## Loglikelihood and Information Criteria:
## 
##   Loglikelihood user model (H0)              -9453.968   -9453.968
##   Scaling correction factor                                  0.996
##       for the MLR correction                                      
##   Loglikelihood unrestricted model (H1)      -9339.126   -9339.126
##   Scaling correction factor                                  1.052
##       for the MLR correction                                      
##                                                                   
##   Akaike (AIC)                               18969.935   18969.935
##   Bayesian (BIC)                             19099.193   19099.193
##   Sample-size adjusted Bayesian (SABIC)      19000.803   19000.803
## 
## Root Mean Square Error of Approximation:
## 
##   RMSEA                                          0.077       0.073
##   90 Percent confidence interval - lower         0.067       0.063
##   90 Percent confidence interval - upper         0.088       0.083
##   P-value H_0: RMSEA <= 0.050                    0.000       0.000
##   P-value H_0: RMSEA >= 0.080                    0.327       0.132
##                                                                   
##   Robust RMSEA                                               0.076
##   90 Percent confidence interval - lower                     0.065
##   90 Percent confidence interval - upper                     0.087
##   P-value H_0: Robust RMSEA <= 0.050                         0.000
##   P-value H_0: Robust RMSEA >= 0.080                         0.277
## 
## Standardized Root Mean Square Residual:
## 
##   SRMR                                           0.050       0.050
## 
## Parameter Estimates:
## 
##   Standard errors                             Sandwich
##   Information bread                           Observed
##   Observed information based on                Hessian
## 
## Latent Variables:
##                    Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
##   SMU =~                                                                
##     sm1               1.000                               0.597    0.547
##     sm2               1.679    0.156   10.761    0.000    1.002    0.760
##     sm3               1.438    0.150    9.582    0.000    0.858    0.685
##     sm4               1.402    0.151    9.265    0.000    0.836    0.528
##   ANX =~                                                                
##     anx1              1.000                               0.920    0.787
##     anx2              0.899    0.073   12.341    0.000    0.827    0.648
##     anx3              1.186    0.063   18.898    0.000    1.091    0.814
##   SE =~                                                                 
##     se1               1.000                               1.018    0.726
##     se2               0.094    0.085    1.104    0.270    0.095    0.091
##     se3               0.689    0.102    6.732    0.000    0.702    0.564
##   WB =~                                                                 
##     wb1               1.000                               0.935    0.718
##     wb2               0.934    0.063   14.817    0.000    0.874    0.699
##     wb3               0.709    0.069   10.211    0.000    0.663    0.456
## 
## Regressions:
##                    Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
##   ANX ~                                                                 
##     SMU               1.312    0.152    8.634    0.000    0.851    0.851
##   SE ~                                                                  
##     SMU               0.984    0.155    6.366    0.000    0.576    0.576
##   WB ~                                                                  
##     SMU              -0.266    0.274   -0.972    0.331   -0.170   -0.170
##     ANX               0.844    0.171    4.930    0.000    0.831    0.831
##     SE                0.340    0.087    3.899    0.000    0.370    0.370
## 
## Variances:
##                    Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
##    .sm1               0.832    0.061   13.669    0.000    0.832    0.700
##    .sm2               0.734    0.079    9.239    0.000    0.734    0.423
##    .sm3               0.835    0.062   13.355    0.000    0.835    0.531
##    .sm4               1.807    0.112   16.199    0.000    1.807    0.721
##    .anx1              0.522    0.052   10.046    0.000    0.522    0.381
##    .anx2              0.946    0.085   11.086    0.000    0.946    0.580
##    .anx3              0.607    0.078    7.748    0.000    0.607    0.338
##    .se1               0.932    0.147    6.318    0.000    0.932    0.473
##    .se2               1.102    0.065   16.926    0.000    1.102    0.992
##    .se3               1.057    0.108    9.820    0.000    1.057    0.682
##    .wb1               0.821    0.075   10.911    0.000    0.821    0.484
##    .wb2               0.800    0.074   10.844    0.000    0.800    0.512
##    .wb3               1.676    0.099   16.998    0.000    1.676    0.792
##     SMU               0.356    0.067    5.342    0.000    1.000    1.000
##    .ANX               0.234    0.052    4.493    0.000    0.276    0.276
##    .SE                0.693    0.143    4.841    0.000    0.668    0.668
##    .WB                0.135    0.062    2.165    0.030    0.155    0.155
## 
## R-Square:
##                    Estimate
##     sm1               0.300
##     sm2               0.577
##     sm3               0.469
##     sm4               0.279
##     anx1              0.619
##     anx2              0.420
##     anx3              0.662
##     se1               0.527
##     se2               0.008
##     se3               0.318
##     wb1               0.516
##     wb2               0.488
##     wb3               0.208
##     ANX               0.724
##     SE                0.332
##     WB                0.845

7.3 Model Fit — Full SEM

sem_fit_idx <- fitMeasures(fit_sem, c(
  "chisq.scaled", "df.scaled", "pvalue.scaled",
  "cfi.robust", "tli.robust",
  "rmsea.robust", "rmsea.ci.lower.robust", "rmsea.ci.upper.robust",
  "srmr"
))

data.frame(
  Indeks    = c("Chi-Square (scaled)", "df", "p-value",
                "CFI (robust)", "TLI (robust)",
                "RMSEA (robust)", "RMSEA CI Lower", "RMSEA CI Upper",
                "SRMR"),
  Nilai     = round(sem_fit_idx, 4),
  Threshold = c("—", "—", "> 0.05",
                "≥ 0.90", "≥ 0.90",
                "< 0.08", "—", "—", "< 0.08"),
  Status    = c("—", "—",
                ifelse(sem_fit_idx["pvalue.scaled"]  > 0.05, "Lulus", "Gagal"),
                ifelse(sem_fit_idx["cfi.robust"]     >= 0.90, "Lulus", "Gagal"),
                ifelse(sem_fit_idx["tli.robust"]     >= 0.90, "Lulus", "Gagal"),
                ifelse(sem_fit_idx["rmsea.robust"]   <  0.08, "Lulus", "Gagal"),
                "—", "—",
                ifelse(sem_fit_idx["srmr"]           <  0.08, "Lulus", "Gagal"))
) %>%
  kable(caption = "Indeks Kesesuaian Model — Full SEM") %>%
  kable_styling(bootstrap_options = c("striped", "hover"), full_width = FALSE)

Indeks Kesesuaian Model — Full SEM
	Indeks	Nilai	Threshold	Status
chisq.scaled	Chi-Square (scaled)	212.4543	—	—
df.scaled	df	60.0000	—	—
pvalue.scaled	p-value	0.0000	> 0.05	Gagal
cfi.robust	CFI (robust)	0.9154	≥ 0.90	Lulus
tli.robust	TLI (robust)	0.8900	≥ 0.90	Gagal
rmsea.robust	RMSEA (robust)	0.0758	< 0.08	Lulus
rmsea.ci.lower.robust	RMSEA CI Lower	0.0649	—	—
rmsea.ci.upper.robust	RMSEA CI Upper	0.0870	—	—
srmr	SRMR	0.0498	< 0.08	Lulus

7.4 Koefisien Jalur (Path Coefficients)

std_sem  <- standardizedsolution(fit_sem)
path_df  <- std_sem[std_sem$op == "~",
                    c("lhs", "rhs", "est.std", "se", "z", "pvalue")]
names(path_df) <- c("Endogen", "Eksogen", "Beta Std", "SE", "z-value", "p-value")

path_df %>%
  mutate(
    across(where(is.numeric), ~ round(.x, 4)),
    Signifikan = case_when(
      `p-value` < 0.001 ~ "*** (p<0.001)",
      `p-value` < 0.010 ~ "**  (p<0.01)",
      `p-value` < 0.050 ~ "*   (p<0.05)",
      TRUE              ~ "ns"
    ),
    Hipotesis = case_when(
      Endogen == "ANX" & Eksogen == "SMU" ~ "H1",
      Endogen == "SE"  & Eksogen == "SMU" ~ "H2",
      Endogen == "WB"  & Eksogen == "SMU" ~ "H3",
      Endogen == "WB"  & Eksogen == "ANX" ~ "H4",
      Endogen == "WB"  & Eksogen == "SE"  ~ "H5",
      TRUE ~ NA_character_
    )
  ) %>%
  select(Hipotesis, everything()) %>%
  kable(caption = "Koefisien Jalur — Full SEM (Standardized)") %>%
  kable_styling(bootstrap_options = c("striped", "hover")) %>%
  column_spec(8, bold = TRUE,
              color = ifelse(path_df[["p-value"]] <= 0.05, "darkgreen", "red")) %>%
  footnote(general = "*** p<0.001 | ** p<0.01 | * p<0.05 | ns = tidak signifikan")

Koefisien Jalur — Full SEM (Standardized)
	Hipotesis	Endogen	Eksogen	Beta Std	SE	z-value	p-value	Signifikan
14	H1	ANX	SMU	0.8506	0.0363	23.4625	0.0000	*** (p<0.001)
15	H2	SE	SMU	0.5763	0.0609	9.4660	0.0000	*** (p<0.001)
16	H3	WB	SMU	-0.1697	0.1696	-1.0001	0.3172	ns
17	H4	WB	ANX	0.8306	0.1517	5.4737	0.0000	*** (p<0.001)
18	H5	WB	SE	0.3704	0.0827	4.4782	0.0000	*** (p<0.001)
Note:
* p<0.001 \| p<0.01 \| * p<0.05 \| ns = tidak signifikan

7.5 R-Square

r2 <- lavInspect(fit_sem, "r2")

data.frame(
  Konstruk = names(r2),
  R_Square = round(r2, 4),
  Kategori = case_when(
    r2 >= 0.67 ~ "Substansial",
    r2 >= 0.33 ~ "Moderat",
    r2 >= 0.19 ~ "Lemah",
    TRUE       ~ "Sangat Lemah"
  )
) %>%
  kable(caption = "R-Square per Konstruk Endogen (Hair et al., 2017)") %>%
  kable_styling(bootstrap_options = c("striped", "hover"), full_width = FALSE) %>%
  column_spec(3, bold = TRUE,
              color = case_when(
                r2 >= 0.67 ~ "darkgreen",
                r2 >= 0.33 ~ "orange",
                TRUE       ~ "red"
              ))

R-Square per Konstruk Endogen (Hair et al., 2017)
	Konstruk	R_Square	Kategori
sm1	sm1	0.2997	Lemah
sm2	sm2	0.5774	Moderat
sm3	sm3	0.4688	Moderat
sm4	sm4	0.2790	Lemah
anx1	anx1	0.6188	Moderat
anx2	anx2	0.4196	Moderat
anx3	anx3	0.6622	Moderat
se1	se1	0.5268	Moderat
se2	se2	0.0082	Sangat Lemah
se3	se3	0.3179	Lemah
wb1	wb1	0.5156	Moderat
wb2	wb2	0.4882	Moderat
wb3	wb3	0.2078	Lemah
ANX	ANX	0.7236	Substansial
SE	SE	0.3321	Moderat
WB	WB	0.8453	Substansial

7.6 Path Diagram — Full SEM

semPaths(fit_sem,
         what           = "std",
         layout         = "tree2",
         rotation       = 2,
         edge.label.cex = 0.78,
         sizeMan        = 5,
         sizeLat        = 9,
         fade           = FALSE,
         edge.color     = "steelblue",
         posCol         = "darkgreen",
         negCol         = "firebrick",
         title          = FALSE,
         mar            = c(2, 3, 2, 3))
title("Full SEM — Social Media Use & Mental Health", line = 1)

8 Pengujian Hipotesis & Interpretasi

8.1 Tabel Rangkuman Hipotesis

std_sem2  <- standardizedsolution(fit_sem)
paths2    <- std_sem2[std_sem2$op == "~", ]

hipotesis_df <- data.frame(
  Hipotesis = case_when(
    paths2$lhs == "ANX" & paths2$rhs == "SMU" ~ "H1",
    paths2$lhs == "SE"  & paths2$rhs == "SMU" ~ "H2",
    paths2$lhs == "WB"  & paths2$rhs == "SMU" ~ "H3",
    paths2$lhs == "WB"  & paths2$rhs == "ANX" ~ "H4",
    paths2$lhs == "WB"  & paths2$rhs == "SE"  ~ "H5",
    TRUE ~ NA_character_
  ),
  Jalur         = c("SMU → ANX", "SMU → SE", "SMU → WB",
                    "ANX → WB", "SE → WB"),
  `Beta Std`    = round(paths2$est.std, 4),
  `p-value`     = round(paths2$pvalue, 4),
  Keputusan     = ifelse(paths2$pvalue <= 0.05, "Diterima", "Ditolak"),
  Interpretasi  = c(
    "Penggunaan medsos berpengaruh terhadap tingkat kecemasan pengguna",
    "Penggunaan medsos berpengaruh terhadap harga diri pengguna",
    "Penggunaan medsos berpengaruh langsung terhadap kesejahteraan psikologis",
    "Kecemasan berpengaruh terhadap kesejahteraan psikologis",
    "Harga diri berpengaruh terhadap kesejahteraan psikologis"
  )
)

hipotesis_df %>%
  kable(caption = "Rangkuman Pengujian Hipotesis (α = 5%)") %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed")) %>%
  column_spec(5, bold = TRUE,
              color = ifelse(hipotesis_df$Keputusan == "Diterima",
                             "darkgreen", "red"))

Rangkuman Pengujian Hipotesis (α = 5%)
Hipotesis	Jalur	Beta.Std	p.value	Keputusan	Interpretasi
H1	SMU → ANX	0.8506	0.0000	Diterima	Penggunaan medsos berpengaruh terhadap tingkat kecemasan pengguna
H2	SMU → SE	0.5763	0.0000	Diterima	Penggunaan medsos berpengaruh terhadap harga diri pengguna
H3	SMU → WB	-0.1697	0.3172	Ditolak	Penggunaan medsos berpengaruh langsung terhadap kesejahteraan psikologis
H4	ANX → WB	0.8306	0.0000	Diterima	Kecemasan berpengaruh terhadap kesejahteraan psikologis
H5	SE → WB	0.3704	0.0000	Diterima	Harga diri berpengaruh terhadap kesejahteraan psikologis

9 Kesimpulan

9.1 Evaluasi Outer Model

Model pengukuran (outer model) dievaluasi melalui tiga kriteria utama:

Validitas konvergen — dievaluasi melalui loading factor (≥ 0.70) dan AVE (≥ 0.50)
Reliabilitas konstruk — dievaluasi melalui Cronbach’s Alpha dan Composite Reliability/Omega (≥ 0.70)
Validitas diskriminan — dievaluasi melalui Fornell-Larcker Criterion (√AVE > korelasi antar konstruk) dan HTMT (< 0.85)

9.2 Evaluasi Inner Model

Model struktural (inner model) dievaluasi menggunakan indeks kesesuaian model (CFI, TLI, RMSEA, SRMR) dan koefisien jalur (path coefficient) dengan uji signifikansi pada α = 5%.

9.3 Temuan Utama

Berdasarkan hasil analisis CB-SEM menggunakan data survei Social Media and Mental Health (n = 478) dengan estimator MLR (Robust Maximum Likelihood):

Hasil keputusan hipotesis selengkapnya dapat dilihat pada tabel pengujian hipotesis di atas.

Temuan ini konsisten dengan Social Comparison Theory (Festinger, 1954) yang menyatakan bahwa paparan terhadap konten media sosial mendorong perilaku membandingkan diri, yang pada gilirannya menurunkan harga diri dan memperburuk kondisi psikologis secara keseluruhan.

9.4 Implikasi

Intensitas penggunaan media sosial yang tidak terkontrol berpotensi memperburuk kondisi kesehatan mental melalui dua jalur utama: peningkatan kecemasan dan penurunan harga diri. Hal ini menggarisbawahi pentingnya literasi digital dan pengelolaan waktu layar (screen time management) sebagai intervensi preventif, khususnya bagi populasi remaja dan dewasa muda.

10 Referensi

Festinger, L. (1954). A Theory of Social Comparison Processes. Human Relations, 7(2), 117–140.
Hair, J. F., et al. (2017). Multivariate Data Analysis (8th ed.). Cengage Learning.
Kline, R. B. (2015). Principles and Practice of Structural Equation Modeling (4th ed.). Guilford Press.
Rosseel, Y. (2012). lavaan: An R Package for Structural Equation Modeling. Journal of Statistical Software, 48(2), 1–36.
Souvikahmed071. (2022). Social Media and Mental Health. Kaggle. https://www.kaggle.com/datasets/souvikahmed071/social-media-and-mental-health

Analisis dilakukan menggunakan R 4.5.2 dengan package lavaan 0.6.21

Structural Equation Modeling

Pengaruh Penggunaan Media Sosial terhadap Kesehatan Mental

Kelompok 2 — M Shidqi Sharim Arrasyid · Anggoro Daru Prasetyo · Frendy Zahril Ramadhani

25 Mei 2026