Reliability Analysis 1 Presented by Shamshuritawati Sharif School
Reliability Analysis 1 Presented by Shamshuritawati Sharif School of Quantitative Sciences College of Arts and Sciences © Shamshuritawati
RELIABILITY TEST 2 Cronbach’s alpha is most commonly used as a reliability measure of a set of items (or statements). This measure can be interpreted as a correlation coefficient, and it’s value ranges from 0 to 1. Items (statements) which are negatively worded must be recoded before performing reliability analysis. The set of items is said to be reliable or have internal consistency if Cronbach’s alpha value is 0. 7 or higher. © Shamshuritawati
RELIABILITY TEST 3 However, Cronbach alpha is quite sensitive to the number of items in the scale. If Cronbach alpha is small, report the mean inter-item correlation for the items. Briggs and Cheek (1986) recommend an optimal range for the inter-item correlation of 0. 2 to 0. 4. If the result is still not promising, carry out factor analysis to isolate multi dimensions or distinct components. Recompute alpha for each component identified. © Shamshuritawati
Example : SAQ. (Item 3 reversed). sav 4 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. Statistics makes me cry My friends will think I'm stupid for not being able to cope with SPSS Standard deviations excite me I dream that Pearson is attacking me with correlation coefficients I don't understand statistics I have little experience of computers All computers hate me I have never been good at mathematics My friends are better at statistics than me Computers are useful only for playing games I did badly at mathematics at school People try to tell you that SPSS makes statistics easier to understand but it doesn't © Shamshuritawati
Example : SAQ. (Item 3 reversed). sav 5 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. I worry that I will cause irreparable damage because of my incompetence with computers Computers have minds of their own and deliberately go wrong whenever I use them Computers are out to get me I weep openly at the mention of central tendency I slip into a coma whenever I see an equation SPSS always crashes when I try to use it Everybody looks at me when I use SPSS I can't sleep for thoughts of eigen vectors I wake up under my duvet thinking that I am trapped under a normal distribution My friends are better at SPSS than I am If I'm good at statistics my friends will think I'm a nerd © Shamshuritawati
RELIABILITY TEST 6 Click Analyze >> Scale >> Reliability Analysis Select “all 23 items” into Items Click on Statistics √ Item √ Scale if item deleted Summaries: √ Correlations Click on Continue Click OK © Shamshuritawati
RELIABILITY TEST 7 Reliability Statistics Cronbach's Alpha Based on Standardized Cronbach's Alpha Items. 806 . 819 N of Items 23 The alpha value is high (> 0. 7). Therefore the 23 items are consistent and thus reliable for measuring student’s anxiety towards learning SPSS. © Shamshuritawati
RELIABILITY TEST Question_01 Question_02 Question_03 Question_04 Question_05 Question_06 Question_07 Question_08 Question_09 Question_10 Question_11 Question_12 Question_13 Question_14 Question_15 Question_16 Question_17 Question_18 Question_19 Question_20 Question_21 Question_22 Question_23 © Shamshuritawati Scale Mean if Item Deleted 59. 89 60. 64 58. 85 59. 48 59. 54 60. 04 59. 34 60. 03 59. 42 59. 99 60. 01 59. 11 59. 82 59. 39 59. 50 59. 39 59. 80 59. 70 59. 97 58. 64 59. 10 59. 38 58. 83 Item-Total Statistics 8 Corrected Scale Variance if Item-Total Item Deleted Correlation 90. 121. 521 101. 064 -. 163 89. 021. 435 87. 968. 569 89. 303. 481 87. 605. 482 85. 656. 594 89. 900. 504 100. 882 -. 137 92. 233. 356 88. 790. 568 88. 452. 563 87. 840. 577 87. 492. 562 88. 766. 484 88. 329. 571 88. 442. 588 85. 993. 609 104. 442 -. 296 91. 699. 314 87. 679. 561 101. 109 -. 153 98. 821 -. 044 Squared Cronbach's Multiple Alpha if Item Correlation Deleted. 373. 792. 188. 820. 398. 794. 385. 788. 291. 792. 427. 791. 470. 785. 490. 792. 220. 829. 197. 799. 530. 789. 424. 789. 451. 788. 393. 788. 344. 792. 463. 789. 494. 788. 492. 785. 209. 832. 270. 801. 454. 788. 167. 824. 086. 819
RELIABILITY TEST 9 Corrected Item-Total Correlation indicate correlation between each item and total score. Low values (<0. 3) indicates the item is measuring something different from the scale as a whole. If it is too low, drop the item. For these data, all data have item-total correlation above 0. 3, which is encouraging. Cronbach Alpha if Item Deleted measures the impact of removing each item. If the values of alpha is higher than overall alpha, it means that the deletion of that item improve reliability. Therefore, we may consider to drop the item. © Shamshuritawati
RELIABILITY TEST 10 Summary Item Statistics Inter-Item Correlations Mean Minimum Maximum Range. 165 -. 342. 629. 971 Maximum / Minimum Variance N of Items -1. 842. 060 23 If Cronbach alpha is high (>0. 7), just ignore the mean for inter-item correlations. If Cronbach alpha is low (<0. 7), check the mean for inter-item correlations and make sure the value is >0. 2. If the value is >0. 2 and Cronbach alpha is <0. 7, we can still conclude that all items were reliable. © Shamshuritawati
RELIABILITY TEST 11 How to report? Cronbach's alpha coefficient for all 23 items is 0. 806. Therefore, it has indicated more than 0. 7 (Nunnally, 1978). All the variables are said to be reliable. © Shamshuritawati
Factor Analysis 12 Presented by Shamshuritawati Sharif School of Quantitative Sciences College of Arts and Sciences © Shamshuritawati
Understanding Factor Analysis 13 Factor analysis is commonly used in: Data reduction Scale development The evaluation of the psychometric quality of a measure, and The assessment of the dimensionality of a set of variables. �Regardless of purpose, factor analysis is used in: � the determination of a small number of factors based on a particular number of inter-related quantitative variables. �The scale must be at least interval. However, in social science studies, Likert scale are often used. © Shamshuritawati
Understanding Factor Analysis 14 � Unlike variables directly measured such as speed, height, weight, etc. , some variables such as egoism, creativity, happiness, religiosity, comfort are not a single measurable entity. � They are constructs that are derived from the measurement of other, directly observable variables. � Constructs are usually defined as unobservable latent variables. e. g. : � motivation/love/hate/care/altruism/anxiety/worry/stress/product quality/physical aptitude/democracy /reliability/power. © Shamshuritawati
Understanding Factor Analysis 15 Generally, the number of factors is much smaller than the number of measures. Therefore, the expectation is that a factor represents a set of measures. � Observed correlations between variables result from their sharing of factors. Example: Correlations between a person’s test scores might be linked to shared factors such as general intelligence, critical thinking and reasoning skills, reading comprehension etc. © Shamshuritawati
Understanding Factor Analysis 16 A major goal of factor analysis is to represent relationships among sets of variables parsimoniously yet keeping factors meaningful. A good factor solution is both simple and interpretable. When factors can be interpreted, new insights are possible. © Shamshuritawati
Application of Factor Analysis 17 � Defining dimensions for an existing measure: In this case the variables to be analyzed are chosen by the initial researcher and not the person conducting the analysis. Factor analysis is performed on a predetermined set of items/scales. Results of factor analysis may not always be satisfactory: � � The items or scales may be poor indicators of the construct or constructs. There may be too few items or scales to represent each underlying dimension. © Shamshuritawati
Application of Factor Analysis 18 � Selecting items or scales to be included in a measure. Factor analysis may be conducted to determine what items or scales should be included and excluded from a measure. Results of the analysis should not be used alone in making decisions of inclusions or exclusions. Decisions should be taken in conjunction with theory and what is known about the construct(s) that the items or scales assess. © Shamshuritawati
Initial Consideration 19 Communalities The communalities for the ith variable are computed by taking the sum of the squared loadings for that variable. Refer Example. Sample size. o Correlation fluctuate from sample to sample, much more so in small sample than in large. Therefore EFA also depends on sample size. o Collect from not < 50, preferably > 100: 20 cases/variable Data Screening Look for at the intercorrelation between variables/items If our test questions/items measure the same underlying construct, then we would expect them to correlate with each other because they are measuring the same thing. © Shamshuritawati
Example : Communalities 20 © Shamshuritawati
Steps in Factor Analysis 21 Factor analysis usually proceeds in four steps: 1 st Step: Correlation matrix for all variables is computed 2 nd Step: Factor extraction 3 rd Step: Factor rotation 4 th Step: Make final decisions about the number of underlying factors © Shamshuritawati
Steps in Factor Analysis: The Correlation Matrix 22 1 st Step: the correlation matrix Generate a correlation matrix for all variables Identify variables not related to other variables If the correlation between variables are small, it is unlikely that they share common factors (variables must be related to each other for the factor model to be appropriate). Think of correlations in absolute value. Correlation coefficients greater than 0. 3 in absolute value are indicative of acceptable correlations. Examine visually the appropriateness of the factor model. © Shamshuritawati
Steps in Factor Analysis: The Correlation Matrix 23 Inter-correlation Correlation matrix : scanning p-value < 0. 05, Correlation matrix: look for multicollinearity (variables highly correlated – R>0. 9) and singularity (perfectly correlated) Determinant: >0. 00001 (no multicollinearity) Anti-image correlation matrix Assess sampling adequacy of each variable MSA<0. 5 is inadequate: exclude the variable Look at the diagonal element of anti-image correlation matrix if KMO is not OK! Ida Rosmini Othman Department of Statistics
Steps in Factor Analysis: The Correlation Matrix 24 Bartlett Test of Sphericity: used to test the hypothesis the correlation matrix is an identity matrix (all diagonal terms are 1 and all off-diagonal terms are 0). If the value of the test statistic for sphericity is large and the associated significance level is small, it is unlikely that the population correlation matrix is an identity. scanning p-value < 0. 05, if so – OK! © Shamshuritawati
Steps in Factor Analysis: The Correlation Matrix 25 The Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy: Measure degree of inter-correlation among variables The closer the KMO measure to 1 indicate a sizeable sampling adequacy (> 0. 9 is superb, 0. 8 and higher are great, 0. 7 is acceptable, 0. 6 is mediocre, less than 0. 5 is unacceptable ). Range from 0. 5 to 1 – Minimum 0. 5 -OK! Reasonably large values are needed for a good factor analysis. Small KMO values indicate that a factor analysis of the variables may not be a good idea. Look at the diagonal element of anti-image correlation matrix if KMO is not OK! © Shamshuritawati
Steps in Factor Analysis: Factor Extraction 26 � 2 nd Step: Factor extraction � The primary objective of this stage is to determine the factors. � Initial decisions can be made here about the number of factors underlying a set of measured variables. � Estimates of initial factors are obtained using Principal components analysis. � The principal components analysis is the most commonly used extraction method. Other factor extraction methods include: � Maximum likelihood method � Principal axis factoring � Alpha method � Unweighted lease squares method � Generalized least square method � Image factoring. © Shamshuritawati
Steps in Factor Analysis: Factor Extraction 27 � In principal components analysis, linear combinations of the observed variables are formed. � The 1 st principal component is the combination that accounts for the largest amount of variance in the sample (1 st extracted factor). � The 2 nd principle component accounts for the next largest amount of variance and is uncorrelated with the first (2 nd extracted factor). � Successive components explain progressively smaller portions of the total sample variance, and all are uncorrelated with each other. © Shamshuritawati
Steps in Factor Analysis: Factor Extraction 28 � To decide on how many factors we need to represent the data, we use 2 statistical criteria: Eigen Values, and The Scree Plot. � The determination of the number of factors is usually done by considering only factors with Eigen values greater than 1. � Factors with a variance less than 1 are no better than a single variable, since each variable is expected to have a variance of 1. © Shamshuritawati Total Variance Explained Extraction Sums of Squared Initial Eigenvalues Loadings % of Cumulativ Variance e% Comp onent Total 1 3. 046 30. 465 2 1. 801 18. 011 48. 476 3 1. 009 10. 091 58. 566 4 . 934 9. 336 67. 902 5 . 840 8. 404 76. 307 6 . 711 7. 107 83. 414 7 . 574 5. 737 89. 151 8 . 440 4. 396 93. 547 9 . 337 3. 368 96. 915 10 . 308 3. 085 100. 000 Extraction Method: Principal Component Analysis.
Steps in Factor Analysis: Factor Extraction 29 � The examination of the Scree plot provides a visual of the total variance associated with each factor. � The steep slope shows the large factors. The gradual trailing off (scree) shows the rest of the factors usually lower than an eigen value of 1. � In choosing the number of factors, in addition to the statistical criteria, one � At this stage, the decision should make initial decisions based on about the number of factors is conceptual and theoretical grounds. not final. © Shamshuritawati
Steps in Factor Analysis: Factor Extraction 30 Kaiser’s criterion Retain factors with eigen values > 1 Scree plot use point of inflexion (find point at which the shape of the curves changes direction and becomes horizontal) retain factors above elbow Parallel Analysis Compare the eigenvalues from FA and simulation using Monte Carlo
Steps in Factor Analysis: Factor Extraction 31 Which Rule? Use Kaiser’s criterion when less than 30 variables & communalities after extraction>0. 7 sample size>250 and mean communality>0. 6 Use Scree plot sample size>250 Use Parallel Analysis to get accurate result and recommended by many journals
Steps in Factor Analysis: Factor Rotation 32 � 3 rd Step: Factor rotation. In this step, factors are rotated. � Un-rotated factors are typically not very interpretable (most factors are correlated with may variables). � Factors are rotated to make them more meaningful and easier to interpret (each variable is associated with a minimal number of factors). � Different rotation methods may result in the identification of somewhat different factors. � © Shamshuritawati
Steps in Factor Analysis: Factor Rotation 33 � The most popular rotational method is Varimax rotations. � Varimax use orthogonal rotations yielding uncorrelated factors/components. � Varimax attempts to minimize the number of variables that have high loadings on a factor. This enhances the interpretability of the factors. © Shamshuritawati
Steps in Factor Analysis: Factor Rotation 34 Other common rotational method used include Oblique rotations which yield correlated factors. Oblique rotations are less frequently used because their results are more difficult to summarize. Other rotational methods include: Quartimax (Orthogonal) Equamax (Orthogonal) Promax (oblique) © Shamshuritawati
Steps in Factor Analysis: Making Final Decisions 35 4 th Step: Making final decisions The final decision about the number of factors to choose is the number of factors for the rotated solution that is most interpretable. To identify factors, group variables that have large loadings for the same factor. Plots of loadings provide a visual for variable clusters. Interpret factors according to the meaning of the variables This decision should be guided by: A priori conceptual beliefs about the number of factors from past research or theory Eigen values computed in step 2. The relative interpretability of rotated solutions computed in step 3. © Shamshuritawati
Assumptions Underlying Factor Analysis 36 Assumption underlying factor analysis include. The measured variables are linearly related to the factors + errors. This assumption is likely to be violated if items limited response scales (two-point response scale like True/False, Right/Wrong items). The data should have a bivariate normal distribution for each pair of variables. Observations are independent. The factor analysis model assumes that variables are determined by common factors and unique factors. All unique factors are assumed to be uncorrelated with each other and with the common factors. © Shamshuritawati
Checklist 37 Based on results. Correlation : scan for p-value > 0. 05, too high coefficient Determinant > 0. 00001 – OK! KMO > 0. 5 and above- OK! , if not check anti-image correlation Bartlett’s test, p-value < 0. 05 – OK! How many factors? Refer Communalities and extraction (all values > 0. 7) Total variance explained (Scree plot) © Shamshuritawati
Factor Analysis via SPSS 38 Presented by Shamshuritawati Sharif School of Quantitative Sciences College of Arts and Sciences © Shamshuritawati
Analyze >> Data Reduction >> Factor 39 1 2 2 1 3 4
Example : SAQ. (Item 3 reversed). sav 40 Fear of computers 6. I have little experience of computers 7. All computers hate me 10. Computers are useful only for playing games 13. I worry that I will cause irreparable damage because of my incompetence with computers 14. Computers are out to get me 15. I weep openly at the mention of central tendency 16. SPSS always crashes when I try to use it 18. I can't sleep for thoughts of eigen vectors Fear of mathematics 8. I have never been good at mathematics 11. I did badly at mathematics at school 17. I slip into a coma whenever I see an equation © Shamshuritawati
Example : SAQ. (Item 3 reversed). sav 41 Fear of statistics 1. Statistics makes me cry 3. Standard deviations excite me 4. I dream that Pearson is attacking me with correlation coefficients 5. I don't understand statistics 12. People try to tell you that SPSS makes statistics easier to understand but it doesn't 14. Computers have minds of their own and deliberately go wrong whenever I use them 21. I wake up under my duvet thinking that I am trapped under a normal distribution Fear of peer evaluation 2. My friends will think I'm stupid for not being able to cope with SPSS 9. My friends are better at statistics than me 19. Everybody looks at me when I use SPSS 20. My friends are better at SPSS than I am 21. If I'm good at statistics my friends will think I'm a nerd © Shamshuritawati
FACTOR ANALYSIS 42 How to report ? : A factor analysis was initially conducted on 23 items with varimax rotation(direct oblimin). However, three items were removed due to cross-loadings. The final model consist of 23 items. The Kaiser-Meyer-Olkin measure verified the sampling adequacy for the analysis, KMO =0. 93 (‘great’ according to Field, 2009), and all MSA values for individual items were larger than 0. 80, which is well above the acceptable limit of 0. 50 (Field, 2009). Bartlett’s test of sphericity 2 (253) = 19334. 49, p-value < 0. 05, indicated that correlations between items are sufficiently large for Factor analysis.
FACTOR ANALYSIS 43 How to report (cont. )? Four factors had eigenvalues over Kaiser’s criterion of 1 and explained 50. 3% of the variance. The scree plot supported the Kaiser’s criterion in retaining four factors. Given the large sample size, and convergence of the scree plot and Kaiser’s criterion on four factors, this is the number of factors that were retained in the final analysis. Table 1 shows the factor loadings. The items that cluster on the same factors suggest that factor 1 represent fear of computer, factor 2 represent fear of mathematics, factor 3 represent fear of statistics and factor 4 represent fear of peer evaluations.
Table 1: Summary of exploratory factor analysis result for xxx questionnaire (N = xxx) 44 Factor Loading Factor 1 Item 2. . Item n Eigenvalue % of variance Cronbach Factor 2 Factor 3 Factor 4
References 45 J. C. Nunnally, Psychometric Theory (2 nd ed. ). New York: Mc. Graw- Hill, 1978 Cortina, J. M. (1993). What is coefficient alpha? An examination of theory and applications. Journal of Applied Psychology, 78, 98 -104. Andy Field Data : http: //www. sagepub. com/field 3 e/Aboutthebook. htm Julie Pallant http: //www. academia. dk/Biologisk. Antropologi/Epidemiologi/PDF/ SPSS_Survival_Manual_Ver 15. pdf http: //www. allenandunwin. com/spss/datafiles. html © Shamshuritawati
Formula 46 © Shamshuritawati
FACTOR ANALYSIS 47 Factor Loadings (Cutoff point at 99% level) – Andy Field Factor Loading Sample Size Needed 0. 722 0. 512 0. 364 50 100 200 0. 298 300 0. 21 600 0. 162 1000
FACTOR ANALYSIS 48 Factor Loadings (Cutoff point at 95% level) – Hair et al Factor Loading Sample Size Needed . 75 50 . 70 60 . 65 70 . 60 85 . 55 100 . 50 120 . 45 150 . 40 200 . 35 250 . 30 350
CFA : Model Fit 49 Fit Indices CMIN p-value CMIN/df Authors Tabachink & Fidell (1996) Marsh & Hocevar (1985); Bentler (1990) Recommended Value Reported: If n between 100– 200 > 0. 05 < 5. 0 Reported: If n > 200 Current Fit Indices 2. 496 RMSEA Byrne (2001); Hu & Bentler (1999) < 0. 08 < 0. 05 0. 089 GFI Chau (1997) ; Segars & Grover (1993) >0. 90 0. 991 CFI NFI TLI Bentler (1990) ; Hatcher (1994) Bentler & Bonett (1980) Kenny D. A (2003) > 0. 90 0. 984 © Shamshuritawati
- Slides: 49