I am in the final year of my PhD and I have used SPSS throughout without any formal training apart from watching your videos. You are the best teacher ever. You are just too phenomenal. You have been a great teacher to some of us low income countries.
I am a PhD student in Epidemiology. My professor just asked me to do factor analysis and I had zero knowledge about this analysis. Your video helped me a lot. Thank you so much.
You are a God sent. I am doing Statistics II and have no clue but through prayer and your videos, I have been getting A's and B's. Thank you.
7 лет назад
You explain everything in a nice manner, no redundant talking as there are in many other videos on youtube. Although I appreciate any effort made, it is really great to watch a video explaining concepts in detail and also being up to the point. Thank you so much!
Great explanation. Way better than my tutors. Please make more of these. You are helping and, reading from the comments, saving a lot of students. as for a comment, It would be nice if you also explained more to WHY you want or need certain numbers it would explain a lot too. KR. Floris
Thank you so so much you saved my thesis,I didnt knew what to do . after watching this video I was able to complete my report. So again a big thank you .
Great tip about using orthogonal rotation when there is no factor correlation > |0.32|. I was reading Tabachnick and Fidell (2007) and came across similar advice about the factor correlation matrix (p. 646).
I just have a question, why we should we have a value of 0.32 or greater in the Component Correlation matrix to choose the Varimax rotation type?, why it is not 0.3 or other values?? is there any source mentioned that?
Clearly an older version of SPSS since that Factor dialogue box looks different in current version. The last part of the video you mention the 0.3 as a threshold for factors and how they were grouped by type. I wish you had spent another 10-20 seconds and elaborated on that a little bit, what that implies exactly. Overall good video and thank you.
i loved your video, it was very informative but i have one difficulty. Does it matter whether the factor loading value is negative or positive or do they both have the weightage. Should we prefer the smaller but positive value such as (0.22) over higher negative value such as (-0.89) when collapsing the variables into factors. Please provide ans to my question
Hi Todd I have two questions. Using different rotations methods, how could you know which one is the best one? component matrix is the loading matrix, isnt it? Thank you very much for your help!
Thank you Sir for the explanation. Sorry i am a little bit confused. I want to do factor analysis precisely CFA for my work, but i don't know if the steps you follow on the video are the same even for EFA or CFA
Dr.Grand, I have some problem in EFA. There are 9 components whicha has eigen value more than one and various items are left with factor loading less than .6 Please help me what to do. KMO is .761 and there is no high correlation above 0.7
Thank you for the helpful video. Two follow-up questions. (1) I read elsewhere the suggestion that the item should be removed if the extraction value (in the Communalities table) is < 0.20. That seems to make sense because too little variance is being explained. Do you agree with this approach? (2) What would you recommend in dealing with items that cross-load across multiple factors in an exploratory factor analysis? Thanks in advance!
Hi Professor Todd. Thank you for the video. It was really very helpful for a beginner like me. A quick question- when we analyze the component Transformation matrix (under direct oblimin) and in the output table we have 2-3 variables whoes value is greater than 0.32, do we still need to change the rotation type to varmax or continue with the old output?? I tried displaying the data with both the rotation types and they are different. What do you suggest??
What is the formal interpretation of having factors load together of it does not mean that the factors are perceived to be measuring same latent variable? or loading together should mean what?
Hello, Thanks so much for this video! I hope you see this question :) I used direct oblimin and Varimax and had absolute values over .32 in both. How do I decide which to stick with? PS: I am doing a factor reduction for over 100 items and both solutions reduced my items to 25. Also my N = 123
Thanks Dr Todd for your very informative videos on Factor Analysis. One question: I notice that the Determinant value for the above Correlation Matrix is 3.81E-5, which to me is 0.0000381 which is less than the thresh-hold value of 0.0001 below which Factor Analysis is not recommended due to high Collinearity, as quoted by various people. Is my understanding incorrect? Thanks again.
dear dr, is there any extraction value that shows problematic commonalities in factor analysis? you said that higher values are better.. right? what low value would be considered bad?
it's very great to learn. I have one question that if I have scores above the .3 on two factors then which and how we can consider that the item xx is on which factor or should be on which factor. thank you
I have a large number of scale items and I need to narrow it down to factors in order to test the hypothesis, so I used factor analysis, I noticed that I can have 3 component, however 1 component when I ran the reliability test for it it was very low (0.4) it is understandable since it only has 2 items, I can't use it because of this. I decided to do what you did as in limit the components to 2 and in the transformation matrix the first component is 0.682 (correlated, I think the word?) with the second, while the second is the negative of that number (-0.682) with the first component and with each other 1 with 1 and 2 with 2 (0.731). Are the values acceptable and can I continue using 2 components? if yes, then how would I interpret the results in my thesis about the component transformation matrix (the one I just mentioned its values) and the total variance explained table (since it shows 3 component not 2)
The level of dissonance I experienced when your true crime video was playing on my tv and this video started playing on my computer as I'm tinkering with spss XD
Thank you very much for the video Dr. TL Todd. I did not however get the point of 'mixture of positive and negative correlation matrix' right. Again you mentioned for symmetrical data SPSS give a KMO value of 0.5 hence your rejection of value as suggestive of appropriate factor. My question is is this true for other softwares? Thank you
What are the differences and implications for extracting factors based on eigenvalues greater than 1 versus using a fixed number of factors to extract while using PAF on spss? Or, do you have any literature that can point me in the right direction? I want to extract a relevant factor for my theory, and when I run my items on spss, using PROMAX, PAF-extraction based on eigenvalues greater than 1, more than one factor extracts from the factor matrix. In prior instances this has happened; even with decent loadings above .500. Is it possible to use a "fixed number of factors to extract," ask SPSS to extract 1 factor only if I can rationalize it through the literature?
Hi , I'm having an issue with combining my questions into 1 and the test the correlation. Should I follow factor analysis method you have shown above or go with computing under transform tab.
I am doing PCA on assets of rural farmers of South Asia. Based on factor loadings on 2-3 different principal components how can we classify them into Poor , Medium and Rich farmers ??
I am not even sure you are going to reply but here goes nothing, I am currently doing my dissertation and my first step on my to do list was to do conduct a factor analysis hoping it would result into three categories but instead it produced over 5 factors so what should I do?
Thank you. I applied factor analysis to data from an employee engagement study and identified 2 factors. How will this result assist me further in identifying and prioritizing interventions?
THank you Dr TOdd! I really understant all easily, what I was looking for! One question please, what is the minumum&maxumum Absolute value below in Options?
Hi, I have a question. I'm testing a model with 3 variables (2 DV, 1 IV). But 1 DV and the IV load onto the same factor since the answers were so similar (Quality of something and the trust). Is this a problem? How should i mention this in my findings. Thanks for all your great videos!
Hey Todd, thank you for this great video! I Just have one question: is this a confirmatory or a exploratory factor analysis? Thank you in advance and best regards, Andrea
Hi Prof. Thanks a lot for your wonderful lecture. I just have one quick question: Is the extraction value for variables the same as what is known as Factor loading?
I have been watching your videos and they have been such a great help! However, I have one doubt. My determinant is actually smaller than 0.00001. How can I proceed? :/
Hi Dr Todd I am still struggling to find unidimensionality through factor analysis. I have been reading a lot of articles and more I am reading more confusing it is getting
sir, why is delete the items that have less correlation of 0.3? if any kind of rules or values? some of researcher saying total number of point values divided by total number of point like agree (1) to disagree (5), .i.e. agree to disagree total number of point is 5 and agree to disagree total value of point is 15 ((agree)1+2+3+4+5(disagree)), therefore 5/15=0.33, so fix the point of correlation is 0.33 it is 5 point scale only. if two and three point scale is change respective of 0.66 and 0.5. 1. 2 point scale is delete the items that have less correlation of .66 with all items. 2. 3 point scale is delete the items that have less correlation of .5 with all items. 3. 4 point scale is delete the items that have less correlation of .4 with all items. 4 5 point scale is delete the items that have less correlation of .33 with all items Show less