SEM with AMOS: From Zero to Hero (20: Structural model assessment)

now we want to access the structure model so this stage is the last stage in structural cushion modeling and we test our hypothesis the relationship between the constructs and consists of three main steps first we access the model fit again because we want to make some changes in the cfa model that we assess this model fit and reliability and validity to convert it to scm model and we need to access the model fit again usually it's very similar i mean if your model fit is good in a structure because i'm in the sem model also you won't face an issue with the model fit but we need to check and report then we estimate the path coefficients we test the hypothesis we compute the p values and we also estimate the r square which is called square multiple correlations so as you already know this is the model that we actually want to test and you have some hypotheses here the hypothesis are basically just the relationship between the constructs in your model for example h1 is there is a positive relationship between subjective norms and attitude then h2 is there is a positive relationship between image and attitude and so on so h1 to h5 um are about the relationship between these five independent variables and attitude and h6 is the relationship between attitude and intention suppose we want to test these hypotheses so what we need to do is we construct a model like this so we don't start drawing the model from the beginning we already fit the model we dropped some items in the cfa yeah this is the cfa model the finalized version so we dropped a few we excluded some items we covariated some of the error terms so we won't change these parts we need these constructs that have good model fit and good reliability and validity and take them to the sem model so what we do is we just save as the model as sem the same model so we don't start from scratch again we don't start from an empty file again we just save as as sem and now we make some changes i first remove the oh sorry we remove the covariances between the it's an attitude and intention and other constructs because only execution of variables means only independent variables should be covariated so what i do is i change the interface properties to landscape a for landscape it's more suitable for my model then i move and put all here um let me move and put in tension as the dependent variable here and this is attitude i just rotate them to yeah make the model more beautiful okay i think it's good and now the next step is we draw the path so all exogenous variables or all independent variables must be covariated as you can see here now you draw a path to follow your research model from all these five to attitude and intention and then from attitude to intention so just when you select a path and in this object then you click in the middle no need to click at the edge then move the mouse then leave the mouse button when you go in the middle of the next one so no need to leave it here at the edge just go in the middle and leave it right um so following the model i call i'll link all these five factors to attitude and intention and the next step is link attitude to intention as well so um let me let's make it a bit more more beautiful select all touch up stick and few clicks and now it's more beautiful and now add the error terms to the endogenous variables so as we aggregate all endogenous variables means all variables that part of their variance is explained by the model means non-independent variables any variable that is depend depends on other variables like attitude that depends on these five intention depends on all these six factors they need error terms you can just double click and give a name to error terms but i use plugins name on all the unobserved variables and gives random names to them um just what to put it here okay um it's a bit out of the screen out of the working space so what i do is i move this a bit to the left now it's fine so this is i saved the file as i used where is the save one as i use the cfa model so no need to link the data file again the data already has been linked and what we do is click on answers properties go to output what you want click here again if you don't know what which one to say select all of them it just gives you more report but in this stage i want actually square multiple correlations as well means r square and what else i want [Music] i would go for okay what i really need is standardized estimates otherwise again you don't get you won't get the standard standardized estimates i need r square and if you want to test indirect effects you select this one but actually i don't need now because the hypotheses are just direct relationships so i will explain i will give you more an example for this one as well too later so basically this is just what we need and modification this is as well so if you ask me this is enough for this stage the first tree and the last one for modification this is i don't see any other thing i want to test now so i close and then i run the model okay i got the red button arrow red arrow button this means i got the results i could run the model successfully and when you click you get the results here on the model so you can switch between on sonder noise and standardized values for example here is 0.22 the relationship between attitude and intention is positive and later we need to check whether it's significant and factor loadings all have been shown here these 36 is r square you see at the corner 63 percent of the variance of intention has been explained by this model this is rs square or a square only on the endogenous variables now it's a bit difficult to read these values um so i click on view text it gives me a report of all things i requested but something that you need to check first is the model fit and model fits based on chi square per degree of freedom based on ifis tli and cfi and ramsey is good usually when your model fit is good in cfa your model fit here should be good as well but we need to report it so you report the model fit indexes those that support your model and of course you report chi squared degree of freedom p-value next step click on estimates you get a few tables here here again you can see the list of tables scholars regression weights is the unstandardized values of the regression weights and what we need in this stage is just this uh because we don't want to check again the factor loadings these are factor loadings right we already checked them in cfa and they are good so what we check is just these rules that are related to the hypothesis the relationship between the constructs right and you can see the unstandardized value for the relationship between subjective norms and attitude is negative but it is not significant this is p-value p-value less than 0.05 it means significant relationship so anything below 0.05 means significant but this is not significant this means this path coefficient is not significantly different from zero so it's actually no significant relationship between these two what is se standard error so estimate divided by standard error will give you critical ratios so if you divide this by this you will get this what is critical ratio it's something like z-score this so basically if it's greater than 1.96 or is less than minus 1.90 in other words if the absolute value of critical ratio is greater than 1.96 this means this path coefficient is significant this means the p-value is less than 0.05 you can see here there are those that are greater than 1.96 the p-value is less than 0.05 what is these three those are these three stars this means the p-value is less than 0.001 so it means very significant right so if here the p-value it means is less than 0.001 and this one is significant this one's significant this one almost significant sometimes i report that's almost significant when it's less than 0.1 significant significant so this one is not significant output quality to intention the direct relationship this one is not significant so the first table is on standardized values the second table here standardized regression will give you standardized values so these are standardized coefficients of the regression waves that you saw in the regression waste table and we already know which one is significant which one is not so if you want to report the standardized regression weights you refer to this table now you may ask which one to report on a standardized or a standardized standardized means they are not dependent on the scale so it will vary between minus one to one so you can compare them right so zero means no relationship when it's close to 1 or this means very strongly relationship and it's close to -1 means very strong negative relationship and this is totally up to you which one you want to report but make sure you keep consistency if you report on a standardized for some of them all must be on standardized and if you want to stand our rewards standard let's report standardize and what else we have here you can get the correlations you can get covariances but what is important is squared multiple correlations the first two are important the rest are just [Music] fact these are just factor loadings the square of factor loadings we don't care about these are about the items so they're just the first two are important the first to attitude intention and we already knew this three five six is here three six it's right we rounded it out up so it's this means 336 percent of the variance of attitude is explained by these five factors and 63 percent of the variance of intention is explained by this model these r squared so um yeah this was actually um structure equation modeling you see so it's very i mean it's compared with cfa is very straightforward and easy so i would say here um let's check the results h1 here subjective norms to attitude is not supported h2 is supported here image job relevance and attitude almost significant almost so up to you how you want to report and yeah you may say non-significant but some people say almost significant to just interpret the findings and output quality to attitude here it is not significant trust is significant and subjective norms to intention okay these five to intention are not our hypothesis in this example so okay i skip then attitude to intention is strong is significant so this it states this is significant so some hypotheses were supported and for some of them the results could not support the hypothesis so now you can report the results of structural model assessment

Channel: saeed sharif

Views: 1,047

Rating: 5 out of 5

Keywords:

Id: elPs4RIK4f4

Length: 12min 55sec (775 seconds)

Published: Wed Mar 17 2021