Model fit reports for possible FAI models 20170313 renew

Model fit reports for ten FAI models

Because the floor effect of our sample (n=420) was marked (23.1%, >20%), we used robust maximum likelihood to correct the estimation bias due to the nonnormal distribution of data.

Criteria for selecting items for each factor in the models that authors did not describe clearly: factor loadings > 0.40 would be included.

*Model 1 Some stroke studies, or most frequently clinical use of FAI
IADL=the sum of item 1~15

Fit statistic	Model fit	Criteria
x2/df	6.58	<3
CFI TLI	0.74 0.70	at least 0.90 at least 0.90
RMSEA SRMR	0.12 0.09	<0.08 <0.08
Conclusion	poor fit

* Model 2 Holbrook & Skilbeck (1983)

Factor 1: Domestic chores: item 1-5

Factor 2: Leisure/work: item 7, 9, 10, 11, 13, 14, 15

Factor 3: Outdoors: item 6, 8, 10, 11, 12

Fit statistic	Model fit	Criteria
x2/df	2.71	<3
CFI TLI	0.93 0.91	at least 0.90 at least 0.90
RMSEA SRMR	0.06 0.05	<0.08 <0.08
Conclusion	acceptable fit

Model 3 Wade et al. (1985) (Delete: No specific naming for one or more factors) 

Factor 1: item 8, 10, 12, 13, 15
Factor 2: item 1-6 (questionable negative high [>0.5]  factor loadings)

Factor 3: item 7, 9, 11, 14 (questionable negative high  [>0.5] factor loadings)

Fit statistic	Model fit	Criteria
x2/df	4.11	<3
CFI TLI	0.86 0.83	at least 0.90 at least 0.90
RMSEA SRMR	0.09 0.07	<0.08 <0.08
Conclusion	poor fit

Model 4 Cockburn et al. (1990) (Delete: no specific naming for one or more factors)

Factor 1 (?): item 1-14

Factor 2 physically demanding activities: item 10, 12, 13

Factor 3 social awareness: item 7, 14 
item 15 (few responses) was not included

Fit statistic	Model fit	Criteria
x2/df	6.73	<3
CFI TLI	0.79 0.72	at least 0.90 at least 0.90
RMSEA SRMR	0.12 0.09	<0.08 <0.08
Conclusion	poor fit

Model 5 Bond et al. (1992)

Factor 1: Indoor domestic chores: item 1, 2, 3, 4, 5

Factor 2: Outdoor social activities: item 6, 7, 8, 11

Factor 3: Outdoor domestic chores: item 10, 12, 13
item 9, 14, 15 were not included due to low factor loadings (<0.4)

Fit statistic	Model fit	Criteria
x2/df	3.49	<3
CFI TLI	0.93 0.90	at least 0.90 at least 0.90
RMSEA SRMR	0.08 0.05	<0.08 <0.08
Conclusion	poor fit

* Model 6 Schuling et al. (1993)

Factor 1: Domestic chores: item 1-5

Factor 2: Outdoors activities: item 6-13

item 14, 15 were not included

Fit statistic	Model fit	Criteria
x2/df	3.64	<3
CFI TLI	0.91 0.89	at least 0.90 at least 0.90
RMSEA SRMR	0.08 0.06	<0.08 <0.08
Conclusion	poor fit

* Model 7 Sveen et al. (1999)

Factor 1: Domestic chore: item 1-5

Factor 2: Outdoor (work/leisure): item 6, 7, 8, 10, 11, 13

Factor 3: Hobby: item 9, 14
item 12, 15 were not included

Fit statistic	Model fit	Criteria
x2/df	3.19	<3
CFI TLI	0.92 0.91	at least 0.90 at least 0.90
RMSEA SRMR	0.07 0.06	<0.08 <0.08
Conclusion	poor fit

* Model 8 Han et al. (2006) ; Lin (2012)

Factor 1: Domestic chores: item 1-6

Factor 2: Work/leisure: item 7-15

Fit statistic	Model fit	Criteria
x2/df	4.13	<3
CFI TLI	0.86 0.83	at least 0.90 at least 0.90
RMSEA SRMR	0.09 0.08	<0.08 <0.08
Conclusion	poor fit

Model 9 Hsieh et al. (2007)
IADL=item 1+item 2+item 3+item 4+item 5+item 6+item7+item 9+item 10+item 11+item 12+item 13+item 15
Item 8, 14 were not included due to misfit to the Rasch model

Fit statistic	Model fit	Criteria
x2/df	7.34	<3
CFI TLI	0.76 0.71	at least 0.90 at least 0.90
RMSEA SRMR	0.12 0.10	<0.08 <0.08
Conclusion	poor fit

Model 10 Han et al. (2009)

Factor 1: Domestic chores: item 1, 2, 3, 4, 6

Factor 2: Outdoor work: item 11, 12, 13, 15

Factor 3: Leisure: item 5, 8, 10

Factor 4: Hobby: item 7, 9, 14

Fit statistic	Model fit	Criteria
x2/df	5.10	<3
CFI TLI	0.82 0.78	at least 0.90 at least 0.90
RMSEA SRMR	0.10 0.08	<0.08 <0.08
Conclusion	poor fit

* Model used in patients with stroke

Conclusion: 
Model 2 had acceptable fit. 

Factor structure of the Frenchay Activities Index in patients with stroke (20170208 renew)

以CFA驗證過去研究所證適用於中風病患的FAI 2-factor model

【多向度Rasch驗證】所得結果

2-factor model 

Factor 1 (domestic chores) = item 1+item 2+item 3+item 4+item 5+item 6
煮飯燒菜、洗碗、洗衣服、簡單的家事、整理家務、商店購物

Factor 2 (work/leisure) =item 7+item 8+item 9+item 10+item 11+item 12+item 13+item 14+item 15 
社交活動、戶外走路超過15分鐘、從事自己的嗜好或興趣、開車或搭公車外出、外出旅遊、照顧花木盆栽、維修汽車或房屋/修理小家電、讀書、工作

Factor 1和Factor 2相關

使用STATA 14進行CFA
配適度檢定結果：did not fit the model

x2/df (需2~5)	RMSEA (需小於0.08)	CFI (需大於0.9)	TLI (需大於0.9)	SRMR (需小於0.08)
5.86	0.11	0.85	0.82	0.08

依據修正指標(modification indices)與期望參數改變量(expected parameter change)修正model為

以下 2-factor model 

Factor 1 (domestic chores) = item 1+item 2+item 3+item 4+item 5

煮飯燒菜、洗碗、洗衣服、簡單的家事、整理家務

其中，
(1) 煮飯燒菜、洗碗 相關
(2) 簡單的家事、整理家務 相關

Factor 2 (work/leisure) =item 6+item 7+item 8+item 9+item 10+item 11+item 12+item 13+item 14+item 15

商店購物、社交活動、戶外走路超過15分鐘、從事自己的嗜好或興趣、開車或搭公車外出、外出旅遊、照顧花木盆栽、維修汽車或房屋/修理小家電、讀書、工作

其中，
(1) 商店購物、戶外走路超過15分鐘 相關
(2) 商店購物、外出旅遊 相關
(3) 社交活動、從事自己的嗜好或興趣 相關
(4) 從事自己的嗜好或興趣、讀書 相關
(5) 開車或搭公車外出、工作 相關
(6) 照顧花木盆栽、維修汽車或房屋/修理小家電 相關
(7) 照顧花木盆栽、讀書 相關

另外，Factor 1和Factor 2相關

配適度檢定結果：fit the model

x2/df (需1~3)	RMSEA (需小於0.08)	CFI (需大於0.9)	TLI (需大於0.9)	SRMR (需小於0.08)
2.94	0.07	0.95	0.93	0.05

結論：

二因子模型：
(1) 家務：煮飯燒菜、洗碗、洗衣服、簡單的家事、整理家務
(2) 工作/休閒：商店購物、社交活動、戶外走路超過15分鐘、從事自己的嗜好或興趣、開車或搭公車外出、外出旅遊、照顧花木盆栽、維修汽車或房屋/修理小家電、讀書、工作
*【商店購物】 隸屬於 【工作/休閒】 因子

臨床使用建議：

以二個分量表的方式使用，各自計分

1. 與家務相關的分量表 (5項目)

2. 與工作/休閒相關的分量表 (10項目)

論文賣點：
1. 首篇以CFA驗證FAI應用於中風病患因素架構的論文
過去FAI應用於中風病患之因素架構驗證：一篇EFA, 一篇PCA, 一篇多向度Rasch
2. 針對最近一篇研究【以多向度Rasch驗證以往研究所提FAI的因素模型，以得適用於中風病患的FAI 因素模型】進行CFA，再次確認適用於中風病患的FAI因素架構，以期提升臨床應用。

FAI CFA結果 / EFA結果

1-factor model

IADL = item 1+item2+item3+item4+item5+item6+ item7+item8+item9+item10+item11+item12+item13+item14+item15

2-factor model

Factor 1 (domestic chores) = item 1+item2+item3+item4+item5+item6

Factor 2 (work/leisure) =item7+item8+item9+item10+item11+item12+item13+item14+item15

Factor 3 (IADL) = Factor 1 + Factor 2

	RMSEA (需小於0.08)	CFI (需大於0.9)	TLI (需大於0.9)	SRMR (需小於0.08)
1-factor model	0.14	0.74	0.69	0.10
2-factor model	0.11	0.85	0.82	0.08

二個model結果都不理想 (2-factor較1-factor好一點)

EFA結果：

Factor 1: item 1 + item 2 + item 3 + item 4 + item 5
Factor 2: item 6 + item 7 + item 8 + item 9 + item 10
Factor 3: item 12 + item 13 + item 14

FAI DIF分析方法與結果(重點紀錄) (20170120 renew)

方法: 
1. 使用ConQuest 2.0軟體分析
2. 以multidimensional PCM分析病患在FAI上每個項目反應

3. 除了確定FAI項目是否符合multidimensional PCM的假設，同時確認有那
    些項目有性別和患側的DIF
4. Item fitting criteria: infit/outfit MNSQ=0.6~1.4
    DIF criteria: difference of logits>0.5 and statistically significant

N=424; M (n=244) 57.5% vs. F (n=180) 42.5%; 
N=388; Left hemi (n=205) 48.3% vs. Right hemi (n=169) 39.9% vs. Bil hemi (n=14) 3.3%

【註】需要重新分析，先確定model fitting，再分析DIF。目前在2-factor model下刪了3題(洗衣服、從事自己的嗜好或興趣、工作)
Domain 1: reliability=0.80
Domain 2: reliability=0.81
有disordering的項目有6個：整理家務、社交活動、開車或搭公車外出、外出旅遊、照顧花木、讀書