Establish Validated Field with airSlate SignNow

Establish validated field

[Music] hello my name is father Rosa and I'm a university professor and John later in this video I will introduce you scale development on validation are you planning to develop your own scale let me guide you through a scan development and how to do it right the process of scale development can be divided into three phases now some even would divide it further into a further scale the phases here in a very simplified way let's talk only about three phases phase one will be about scale generation above the items and the scales how to develop and generate them in the second phase we would look into refining them and then finally in the third phase would validate the scale once then you've completed that process the scale is ready for use in your research projects starting with phase one the objective of phase one is to define your construct and you define and identify the dimension or if you have a multi-dimensional scale your demap dimensions and maybe sub dimensions that would be usually based on literature you would carefully review this excellent literature be very clear about your your construct ideally also there should be theory behind it should be theory driven in the next step we would maybe go further to literature not just looking at the overall construct but maybe also in the very specific items maybe some related scales exist or maybe some related ideas exist it could be also conceptual or qualitative work and then it's about you to develop a scale to measure that construct furthermore you will also like to validate your items and your scales in terms of external validity what it is and how it could be measured and you would generate items but you will already maybe have some experts and there could be people who about that Omega odds your target population and if you're studying expatriates you will may be serving expatriates not survey you would interview and ask them about certain constructs and items whether they make sense to them you could also add some other experts other maybe researchers are the consultants or people who know about it you know and so in this face some what you have two sub phases the first part is defining constructs identifying dimensions and then the second part really go to the items you try to generate a list of items and already review the items through expert interviews that concludes then phase number one the question then is that I sometimes years how many items do I need for one dimension for one scale I think we have to think backwards for me an ideal number of items of our scale would be three or let's say three to five because you need three to have a nice variation and in terms of your scale but in order to derive at three items in your final scale you would start with a much larger number of items it's hard to say based on my own experience you can divide the number of initial item pool maybe by two or maybe by three so meaning here this example the earthen face one maybe you will start with eight items in Phase two well then maybe you have to drop items in Phase three maybe of to drop more items so maybe at the very initial stage maybe if you would like to end up with three items maybe it's better to start with eight items at the beginning and then it will shrink throughout this process now coming to phase number two it's about scale refinement what you would typically do is you would run an exploratory effect analysis the objective is to check your dimension if you've proposed three dimensions you would like to test through an exploit or NASA's whether indeed listeners identifies three dimensions and then maybe also member your items load on your proposed dimension and not on other dimensions typically we would run pilot study one on a sample of two to three hundred respondents and then run this exploratory effectiveness and check the dimensionality and reduce items it's very common and that we drop items during this process what item we would drop items if the factor loading on this respective factor is rather low maybe below point four below point five and or if the item has a high cross loading with another dimension that is is not supposed to load on at this stage also check for reliability or called bus alpha it should be at least above point seven now imagine we have completed the Phase two we can move on to Phase three typically what we will do is then we collect another sample pilot study to maybe at least 200 or maybe 300 respondents will participate we have now a reduced number of items maybe even reduce some of dimensions and we will then provide this list of items to our new sample to evaluate the scale in contrast to exploratory effect analysis before there are all items into their respective dimensions typical effectors could look like this one we have maybe three dimensions each with different items and then we would look in the model faith good look into the factor loadings whether each item Lawson's respective factor and then we can determine the model there's some very rough cut off criteria that has been proposed here I've listed some of the most important ones in addition of course he would always report chi-square degrees of freedom and the many other criteria but the most common ones would be probably CFI and tli and so these indicators we would usually say they should be at least point nine or higher that would be the loose criteria the more stricter one will be higher in point nine five and then we've our our Mercier and then it should be below point or eight or below point or six if we use use the more stricter number if you developing a scale I would usually recommend to apply more stricter rules here because if you would like to apply your scale as well you have a different sample and so forth it's not perfectly fitted to the sample then maybe also the model fit and all these indicators would be lower so it would be safer to be more critical here this development stage also there's the question now your model does not meet these criteria well model fit is much lower CF is much lower and so forth what do you do if you developing a new scale and you will first look into factor loadings error correlations maybe modification indices and so forth if you're doing scale development you would usually drop items that's also why I recommend you should have more items to begin with so you can drop them later on if you're not developing a scale but you're just fella dating you've used another scale and now we apply them in your research context but maybe you can also think about air or college connecting certain error terms if it is justified you would look into the wording if the wording is somewhat similar and you would do so but for scale development purpose rather you drop items that would for scale validation is extremely important to establish discriminant validity meaning to show that your variables are distinct distinct from one another in particularly it would be extremely important to show that your independent variable is distinct from your dependent variable meaning they're not the same there's not much overlap the different ways how to do so it's also somewhat field-specific in management research organizational behavior research it is very common to conduct multiple confirmatory effect analysis you would compare different confirming the results of confirming effect analyst and you would use different numbers of variables you may combine certain variables maybe the original model has six vectors and then you would run the five factor model where you combine may be independent dependent variable or you would combine different variables you will compare the full model with maybe reduce factors where you combine certain variables and then you also want to combine all variables and then you will need to show that the model fit of the models with fewer with fewer variables is inferior to the full model or if it's not then you have a problem of this grand validity another method that has been very popular is the phone and Locker method you would look into average Barretts extracted and composite reliabilities and then you would compare the squared the square root of average average variance extracted with the other correlations and then they should be lower and then you would also conclude you have this credibility another method not so common these days we'll be looking at cross loadings the items it's a level that the items should be well correlated with the same with the items in the same vector but they should be a much less correlated with items from other factors in recent years also we have the new methods its interest rate monitor age ratio of correlations or HTM T and it looks also items and again like items with correlations within its factors compared to a correlation with items outside the original proposed effectors and they compute ratio this has been proposed by a group of authors more use and information systems research and maybe marketing at the bottom you find a reference to an article that explains HTM T and also I like this article because it also introduces the other ways of how to calculate this current validity in this video will not go into details of these different ways of how to establish this going me I just explained and if you're interested I refer you to their excellent literature and also other videos in phase 3 in addition to scaley evaluation through confirmatory factor analysis you would like to establish a new more logical network so that your new scale should be related to certain variables that would we expect and if they should be unrelated with some other variables to do so you would usually collect another sample may be pilot study 3 where you include related and unrelated pheromones but maybe you would run correlations or other tests to show overlap and non overlap with those variables if you have gone through all these three phases then I can congratulation you have established a new scale I think you get the picture scary bellman is very complex that's you may want to think twice rather to develop a new scale alternatively you can also use an existing scale and maybe modify adapted to your context that would be much easier and you don't need to go through all these different steps if you want to do it right I describe these three steps first scale generation second scale refinement and third scale validation I hope this introduction gave you an idea of what it means the event of the new scale good luck bye bye