How to compute composite variables in SPSS: Examples using (fictional) survey and performance data

hi everyone in this video i wanted to spend a little time showing you how to compute composite variables in spss and we will be using some fictional survey uh and performance data so before we get started i want to mention that underneath the video description you will find a link to the spss data file that i'm working from throughout this demonstration additionally you will find a link to a powerpoint and that powerpoint is a very short one but contains some of the information that i'll be covering so let's go ahead and get started by opening up our spss data file and looking at its contents so here we have our data file opened up and basically the first uh 10 variables in the data set are reflecting uh the items within our survey the first five items are essentially reflecting a person's are designed to measure a person's interest in politics the next five items are designed to measure their self-efficacy for learning politics and then at the end of the data set we have three performance measures so in terms of their political knowledge comprehension and reasoning about political issues now the numbers that are shown as they're associated with each item represents a person's responses on either the survey with respect to the interest and efficacy items or responses or basically scale scores with respect to the performance measures right here so before we start scoring our our composite measures or creating our composite measures just really quickly i wanted to show you sort of a factor analytic representation of what the items are designed to measure so you can see that our interest items we have interest one through interest five these are all supposed to be measuring a late construct called political interest so the reason why the errors are pointing to each of these boxes right here is because the idea is that when individuals are responding to the interest items at least part of the variation should be attributable and hopefully most of the is attributable to the political interest factor the e's over here just represent error variances but we're not going to get into all that you also see with our political competence measure we have our knowledge comprehension and reasoning scores if you will and so the idea is that how people respond to these sort of scale scores would reflect their level of political competence and then over here we have self-efficacy our factor being indicated by our efficacy items within the scale so basically what happens in in applied research is we you know we'll take the individual measures of the construct and we will aggregate them in some way maybe we sum them up or average them out but in the end the idea is to come up with a full scale score that represents where a person falls on that particular construct now the downside in a lot of this is that we are not able to tease out that measurement error associated with our uh scales and so it's for that reason that we want to make sure that we report at least on the reliability associated with any kind of multi-item or indicator of measure so let's go ahead and take a look at our items and then we'll begin to compute our scale scores so with respect to the interest and fc items you'll see that the first five items in our measure here um are about our interest in politics and all those are worded in the affirmative so basically if a person indicates uh that they strongly agree disagree by uh replying with a one um they're basically indicating for all of these items that they are not interested in politics whereas if if they indicate a five which would be a strongly agreed then that indicates that they are uh very interested in uh politics now with respect to the next five items right here uh we actually have a couple of sort of negatively worded items so you'll see with item six it says i feel confident in my ability to understand political issues seven i can successfully learn more about political issues eight is a negatively worded item because it says i do not understand what people are talking about when they discuss political issues and the same goes with item 10 i am never going to be able to understand what politicians are talking about so the thing is is that if we're using the scale from one to five uh it makes sense with six and seven uh because in both of those cases a one would indicate that they're not interested in political issues uh five would indicate that they're very interested but on item number eight and item number 10 if they indicate strong agreement then what that actually is indicating that they strongly agree with the opposite sentiment that's reflected in the remaining item so what that's going to mean is is that when we compute our scale score we're going to want to make sure to reverse code those two items or or else it's not going to make much sense with our with our full scale score so let's go ahead and open up spss and begin with computing our our full-scale scores or composite scores okay so with our interest items right here we don't have to do any kind of reverse coding because all of the items are word worded in the same direction so what i'm going to do is go up to the transform button right here and then go to compute variable and under here i'm just going to create a target variable so i'll just call this uh interest or yeah that work and then under numeric expression we can do this in a couple of ways if you want just to sum up the items you can actually do that under numeric expression pretty easily you can just you can either type or you can move things over here you can just say interest 1 plus interest 2 plus interest 3 plus interest four plus interest five and when we click okay right here um at the end of the data set you'll see that we now have our interest variable uh that's given so we basically have summed uh each individual's responses across those uh five items um another option if you want to do this a little bit quicker you can you can do it this way you could just say uh sum and then you can just say interest one comma interest 2 comma interest 3 comma interest 4 comma interest 5 and that will work too you don't have to type in i'm just going to save over that you don't have to type in each of the plus signs so you can see it's summed everything right there if you want to compute the mean of those items if that was your preference you can do the same in fact i'll just create a new variable i'll just call interest mean right here and instead of using the sum function right here we can just type mean and then inside the parenthesis we still have our items that are separated by the commas and we can click ok right here and so now at the end of the data set we've got that variable uh that's given so uh the difference is is that with the interest um you know the the range of of um of uh values on this variable uh is not going to be between say one and five so if you wanted to kind of talk about the average response to the items uh for each person you could use this variable right here kind of the interest mean to kind of capture the average response for of a given individual across those items but you can't really talk about it as cleanly with that interest variable right there but you know just keep in mind that basically they're they're both uh representing the composite measure for the interest items and in fact if i you know if i take these and i correlate them both you're going to find that they're going to correlate at one so you know they're just kind of different ways of expressing uh the same idea which is uh where a person falls on the composite measure now with respect to the fc items we saw that i am basically the item one and two and item 4 all of those are are positively worded whereas items 3 and item 5 which were the 8 and the 10 in the powerpoint those two items are reverse coded and so what we're going to do what we're going to want to do first is to reverse code these two items so that when we sum them up we're all you know we're getting an expression uh where higher scores represent greater efficacy and lower scores represent less efficacy and we want that to be reflected in the same way across those items so what we can do is to use the recode function first before we compute our composite scale so if we go to transform you actually have a couple of options you've got recode in the same variables and recode into different variables and i'll be honest i i tend to prefer to go with the recode and the different variables because it kind of leaves a trail if i'm working with a large data set and and going through lots of different analyses it's nice to have some breadcrumbs so to speak to go back and figure out what you've done up to this point so i'm going to use the recode into different variables function and i'll show you what i'm what i'm talking about here so in this case we're going to take uh item uh 3 and we're going to move it over to this in this box right here we'll also do the same thing with item 5 and we'll move that over here as well so what i'm going to do is i'm going to create two new variables in the data set that have been recoded so i can move the go up here to fxe3 and i'll just type in fc 3 and i'll put a little r right there just to indicate that this is a recoded item i'll click change and then we'll do the same thing for fc5 so i'll type in fc 5 and then lar and then followed up with clicking change right there you can name these whatever you want to this is just kind of how i tend to organize things when i'm uh setting up um data sets and and recoding and so forth so i just it's just how i do it i tend to use a little r to designate that now at this point what we're going to need to do is to go under old and new values so here we've got old value i'm going to type 1 and new value is going to be a 5. so you know remember those items where a 1 on those items basically a strong disagreement with an item that is negatively worded is actually reflecting greater efficacy so i want to convert the 1 to a 5 right here and i'll click add then i'm going to do a 2 and a 4 right here i don't really need to do anything with the 3 because that's the midpoint of the scale but i'll go ahead and do that sometimes i like to do this just as a little placeholder and then we'll do a 4 and a 2 and then a 5 and a 1. and click add right here and from there when i click continue and then on ok you'll see that now in my data set i've got these two reverse coded items so there's fc3r and fxe5r and i know that it's showing up as a nominal variable but uh it's not going to make any difference in terms of when we're computing things um but if you wanted to change it you certainly could we could just go over here and just change the scale to or change the the scale uh to a scale variable in spss um so at any rate now i can sum up uh items 1 2 3r and 5r into a four r four and four five are into my composite fc measure so what i'll do is i'll go to transform compute variables and now we'll uh go ahead and type in for a target variable i'll just call this efficacy tote if you will and again if i wanted to sum this up i can use the sum function and then then move each of these variables over so i can kind of double click right there comma double click comma then move to 3r down here comma then 4 comma and then 5r right there and in parenthesis and so when i click ok you can see that now that variable is showing up and once again if i wanted to use the sort of average response as the full scale score the average response for an individual across the items i could do that very easily too if i go back to compute variable i'll just type in right here i'll just say avg for average if i want and instead right here we will just type in mean and then when i click ok right there you can see that now i've got my fcc total full scale and then the fc average so basically those are just some common things that you might use when you are computing full scale scores using survey items and so now let's just really quickly uh do it with our political knowledge comprehension and and reasoning measures there's not going to be any kind of recoding that we're going to do we're just going to go to transform compute and i'm going to press reset right here and then we'll just call this i'll just call this competence and then under the numeric expression i'll just go ahead and type in mean right here and and then we'll take our uh political knowledge comma uh political comprehension comma and then political reasoning uh variable and then we have our in parenthesis click ok and so now you can see i've got the a composite measure of my competence indicators okay so um at any rate that pretty well concludes this demonstration of how you can compute composite variables using spss and i appreciate you watching

Channel: Mike Crowson

Views: 2,294

Rating: 4.909091 out of 5

Keywords: Computing scale scores using SPSS, Recoding survey items in SPSS

Id: 9jj-oPcu23M

Length: 15min 4sec (904 seconds)

Published: Fri Jan 22 2021