This sounds like an interesting study. If you posted the results back here I think we'd be interested in seeing your results. Anyway on to your question:
How did you measure this?desire of members to share information
Hi guys! Please i need your advice!
I have developed a model about some factors that influence the desire of members to share information is online communities. So, i have in total 9 variables and I want to check the relationship among the variables.
My question is which of the following method is the best choice:
a) linear regression
b) confirmatory factor analysis
I have already run the the linear regression in SPSS and the results made sense. So what is your opinion?
Thanks in advance!
This sounds like an interesting study. If you posted the results back here I think we'd be interested in seeing your results. Anyway on to your question:
How did you measure this?desire of members to share information
"If you torture the data long enough it will eventually confess."
-Ronald Harry Coase -
Hi! For all the variable I asked two or three questions. So regarding the "desire of members to share information" i had some question like the following one:
a) I visit the yyy community to share my knowledge with other people
about the answer i used the five item scale that ranges from strongly disagree to strongly agree...
So do you have any idea about the research method?
I agree. This sounds very interesting. You mentioned that you want to test the relationship between variables. Does this mean you want to test this between your nine variables OR do you wish to understand the independent effect of the nine variables on desire to share information?
The relationships that i want to study are the followings:
X1--> Y1
X2--> Y1
X3-->Y1
X4-->Y2
X5-->Y2
Y1-->Y2
Y1-->Y3
Y2-->Y3
So, Y3 is the the desire of members to share information, let's say the main dependent variable
X1...X5 are the independent variables
what is X1, X2,... X5 and Y1, Y2, Y3?? this is sounding a lot up the alley of either path analysis or structural equation modeling..
X1...X5
and Y1... Y3 are the variables of the model
yah, i got that from your previous post, but i need (actually to give you proper advice we all need) to know what they are. for instance i know Y3 = desire of members to share information because you said so. what is Y1? what is Y2? what are X1, X2,...X5? age? gender? motivation? depending on whether you have purely observed variables or latent variables with indicators, this thing could either be some type of regression, path analysis or structural equation modeling... it looks more like structural equation modeling because you mentioned confirmatory factor analysis (which happens to be a special case of SEM) but, to be able to say that, more info about your research design is needed...
thanks!
Hi!!!
Yes of course i can share this info with you.
Y1=Members' satisfaction
Y2= Sense of Belonging
Y3= Members' desire to share information
X1= System's Quality
X2= Information Quality
X3= Ease of use
X4= Trust
X5= Length of membership
I have already run the linear regression in SPSS, and the results make sense since i have calculated r, r square, adjusted r, t-statistic, sig and the Beta. So is this enough to verify the validity and the correlation among the variables, since i know that the linear regression does not prove the existence of a cause-effect relationship...
Is it necessary to use Lisrel for a factor analysis?
Thanks in advance!
uhm... this is where it gets tricky and it depends on how sophisticated you'd like to get. mathematically speaking that should suffice because you're finding evidence of linear realtionships among variables... the problem is that, if you are building this as an explanatory model for, how did you say it? measure the desire of members to share information is online communities then the answer would be "no", because desire to share information is a latent variable in a measurement-validity sense which relates to a psychological construct and OLS regression only deals with directly observable variables. you would be making the assumption that whatever score people get in your Y1, Y2, Y3 variables is in fact a direct measure of members satisfaction, sense of belonging and desire to share information... and we know that is not the case because you cannot cut a piece of desire to share information and weight it or measure it in inches to see how much of it is present in an individual... in that case you need to tap into the variance of the latent variable through your indicators (your X's) and build a structural model on which you do latent variable regression...
ok, so... what is this particular piece of research for? is it for a school paper i think going the way of OLS regression should suffice, but if this is something you may want to end up publishing at some point, you'd need to go and do SEM...
(oh, and LISREL is not the only piece of software that does confirmatory factor analysis. you can do that in R, SAS, Mplus or the optional/additional package AMOS from the guys who created SPSS...)
@ Spunky
It is about a master thesis assignment so from your point of view it is better to so SEM by using one of the tools that you mentioned in your post right?
?
master thesis assignment? then yes, you'd wanna use SEM... absolutely. if you already know that there is such a thing as confirmatory factor analysis and LISREL and all, then you're expected to use SEM... besides, (i was just TAing for a SEM-like course) i forgot to mention that you'll need the simultaneous solution to all the regression equations you're specifying and SEM is the only way to get those...
However, for SEM (Structural Equation Modeling) there are three methodologies that can be used:
a) Regression (both linear and multiple)
b) Confirmatory factor analysis
c) Path analysis
ok... regression is in itself not structural equation modeling but it's a building block to it. however, you have several independent and dependent variables which OLS regression cannot handle because it is a univariate (only one dependent Y) technique. OLS regression also assumes that you're measuring the variables you're regressing directly and without error, which is not your case because you have both observable and latent variables. so regular regression is too limited for what you'd like to do...
now, the choice is now between confirmatory factor analysis or path analysis and that kind of depends on the kind of data you have. confirmatory factor analysis is a specific application of structural equation modeling to test validation... if your Xs and Ys are are items for psychological scales/instruments and your Xs and Ys are the latent variables on which you are assuming they'd load, then you use it... however, it seems to me that you're specifying a strucutral model on your data with a whole bunch of observable indicators tapping into the latent variables... so i think you should go for path analysis, which is the most general version of SEM...
Thanks a lot i will finally apply the confirmatory factor analysis even though i have not understood yet its functionality. In the literature most of the researchers have used CFA (Confirmatory Factor Analysis)
Tweet |