Purpose: to examine the construct validity of the All Aspects of Health
Literacy Scale (AAHLS)
Measure: The AAHLS comprises 13 items covering 3 subdomains: functional
HL, communicative HL, and critical HL.
Data: 病人與家屬的data合併分析 (n=139)所得結果較僅分析病人的data (n=107)好,故以下呈現病人與家屬的data分析結果
Data analysis and critical values:
1.
CFA: χ2/df=2, RMSEA=0.06,
CFI=0.95, SRMR=0.08
2.
Rasch (PCM) analysis: infit and outfit MNSQ = 0.6-1.4
3.
Mokken scaling: Hi ≥ 0.3, scale H ≥ 0.5
method
|
unidimension
|
three
dimension
|
CFA
|
initial model:
χ2=221.28 df=65 χ2/df=3.40 RMSEA=0.13 CFI=0.73 SRMR=0.09 ※ Poor model fit!
刪除factor loading <0.5的5題(第1-3、11、13題):
χ2=121.64 df=20 χ2/df=6.08 RMSEA=0.19 CFI=0.80 SRMR=0.08 ※ Poorer model fit! |
initial model:
χ2=117.24 df=62 χ2/df=1.89 RMSEA=0.08 CFI=0.90 SRMR=0.07 ※ Problems:第1題的error variance為負值,不合理,代表model有問題;若刪除第1題,換第2題的error variance為負值,然若同時刪除第1、2題,functional HL subdomain僅剩一題,無法以CFA分析 (model unidentified)。故刪除functional HL subdomain重新分析
functional HL subdomain deleted:
χ2=70.20 df=34 χ2/df=2.06 RMSEA=0.09 CFI=0.93 SRMR=0.06 ※ RMSEA value was quite large, indicating inadequate fit, although the values of the other indices were acceptable.
刪除factor loading <0.5的2題(第11、13題):
χ2=58.65 df=19 χ2/df=3.09 RMSEA=0.12 CFI=0.92 SRMR=0.06 ※ poorer model fit! |
Rasch
|
5 misfit
items were deleted (items 1-3, 11, 13):
PCA 85.5%
Rasch
reliability=0.77
|
3 misfit
items were deleted (items 10, 11, 13):
Rasch reliability:
functional HL=0.45 communicative HL=0.80 critical HL=0.82 |
Mokken
|
刪除5題Hi < 0.3的題目(items 1-3, 11, 13):
scale H=0.55
Reliability RHO= 0.89
|
|
後續論文撰寫計畫:分析結果不佳,若寫成論文,文章難以凸顯重點,發表機會不高。此外,樣本數不夠大,無法有力拒絕此工具之建構效度不佳。故資料暫擱置,留待日後發展測驗時之參考。