Interpreting SPSS Output for Factor Analysis

hello this is dr. Grande welcome to my video on interpreting the SPSS output for a factor analysis I have here in SPSS in the data editor view fictitious data and I have 10 variables item 1 all the way through item 10 so these variables represent individual questions or items on a psychometric instrument and let's assume that we have a reason to believe that item 1 2 3 & 4 are related to one another item 5 6 & 7 are related and item 8 9 and 10 are related so we believe that in these data we would find three factors that these items would load cleanly on three factors so if we conduct a factor analysis we would expect to be able to interpret the rotated component matrix and see that item 1 2 3 & 4 load together item 5 6 7 did as well and 8 9 10 loaded together but even if we had that as a finding that wouldn't mean that these items are measuring the construct for which the psychometric instrument was designed the factor analysis will tell us if the items load together but it does not answer the question of what construct is being measured so to conduct a factor analysis here we're going to go to analyze dimension reduction then factor this is what the factor analysis dialog looks like by default over here in the left you can see item 1 through item 10 I'm just going to hit ctrl a and select all these and then move them over to the variables list box under descriptives I'm going to add the univariate descriptive z' and under correlation matrix the coefficients significance levels the determinant and kmo and Bartlett's test of sarisa t click continue under the extraction item I'm going to add the scree plot I'm going to leave unrotated factor solution checked off and I'm going to leave the based on eigen value option set at the default which is I ghen values greater than 1 are extracted click continue and then under rotation I'm going to need to determine the rotation I can use so I'm going to start with direct of lemon and this is an oblique rotation and we will interpret the output and see if we we're going to stick with oblique and oblique rotation like the director of lemon or Promax or if we can move to orthogonal rotation like Vera Mac's quarter Max or aqua Max and I'm going to leave the rotated solution checked off and click continue I'm not going to make any changes underscores and I'm not going to make any changes under options now but I'm going to come back and make some changes here later to show you how it changes the output so I'm going to click OK and run this factor analysis as indicated by the parameters that were selected you can see we have the descriptive statistics the correlation matrix and a great deal of other output but for this particular run of the factor analysis I'm only interested in the component correlation matrix because first I want to determine what rotation method I should be using what I'm looking for here in this correlation matrix you can see I have three factors or three components I'm looking for a value here a correlation value that when I take the absolute value of the correlation value it's greater than 0.3 - and you can see in no case here other than the factor correlating with itself which of course is going to be 1 and no other case do I have a value whose absolute value exceeds 0.32 so in this case I'm going to go back to analyze dimension reduction factor and I'm going to change the rotation to Vera max if it was greater than 0.3 - if the at any of those values if the absolute value is greater than 0.3 - I would stick with director of lemon or pro max in this case I'm going to move to Vera max and click continue and then click ok so now to interpret the output from the factor analysis and you can see the first table we have descriptive statistics and of course we have the mean and the standard deviation for all the items and then we have the sample size that's the analysis n you can see it's 200 and I use this sample size because generally 200 or above is considered acceptable and below 200 is considered poor so generally 200 would be the fewest number of records that you would want to analyze then we have a look at the correlation matrix you can see that we have all 10 items correlated with the 10 items and of course diagonally we'll have all perfect correlations so we're provided with the correlation values and then we have the p-value here in this table below the next table is kmo and Bartlett's test you see the first item is the kaiser meyer hogan measure of sampling adequacy and the value we have here is 0.72 - in general anything above 0.5 is considered acceptable although a value above 0.6 is preferred as far as Bartlett's test of cerissa tea we look at the p-value here we have Oh point zero zero zero which normally we would record as less than point zero zero one in this case that's the result we want we want to have a statistically significant value for Bartlett's test of sarisa tea so a value below 0.05 so next we have the communalities table and you can see for initial all the values are 1 and we have some different values here for extraction this extraction value tells us the proportion of variance for each variable that can be explained by the factors so in this case looking at these extraction values they're very high so these are good extraction values then moving down to the next table we have total variance explained and we can see that SPSS extracted three factors or components and the cumulative percentage was 85 point eight five one so these three factors explained 85 percent of the variance and if we move down to the scree plot we can see that three factors were above an eigen value of one and all the other potential factors were below that so they were not extracted then moving down we have the component matrix and the rotated component matrix and this is a table that's easier to interpret when you make some changes under options in the factor analysis dialog I'm going to go back and make those changes so that this becomes easy to read but even without the changes I'm going to make you can see that factor one has item one thread and for very strong loadings and then we have five six and seven again very strong loadings at eight nine ten the factor loadings are high now I'm going to move back to analyze dimension reduction factor and of course it's going to retain all the properties from this analysis so I'm just going to go into options I'm going to sort by size for the coefficient display format and I'm going to suppress small coefficients now there are several theories about what a factor loading minimum value should be one of the popular values is 0.3 so I'm going to use that an absolute valuable 0.3 those factor loadings will not be displayed click continue and click OK I'm going to move back down to the rotated component matrix you can see that the factor loadings are sorted by size so 0.98 1 is the strongest factor loading so it's up top and point 8 3 4 is the weakest so it's at the bottom make see we have item 1 2 3 & 4 all load together item 8 9 and 10 although together and then 5 6 & 7 all load together so by not displaying the factor loadings that are below what we would consider to be a significant loading value and by sorting these by size the rotated component matrix becomes easier to interpret I hope you found this video on interpreting SPSS output for factor analysis to be helpful as always if you have any questions or concerns feel free to contact me and I'll be happy to assist you

Channel: Dr. Todd Grande

Views: 272,799

Rating: 4.8558187 out of 5

Keywords: SPSS, factor analysis, rotation, rotations, rotated component matrix, simple structure, complex variable, KMO, Bartlett, Sphericity, communalities, KMO and Bartlett’s Test, total variance explained, factor loading, factor loadings, significant, zero loading, orthogonal, oblique, varimax, quartimax, equamax direct oblimin, promax, factor analysis tables, descriptive statistics, correlation matrix, component matrix, matrix, counseling, Grande

Id: g_3kaSnq-DY

Length: 10min 41sec (641 seconds)

Published: Thu Mar 17 2016