COMM 301: Empirical Research in CommunicationStatistics: Combining items into scales on SPSS and testing scale reliabilityIf your research uses Likert or semantic differential scale items to survey attitudes or opinions, there are a series of data manipulation procedures you have to carry out before you can run further analysis. It involves - recoding variables,- testing the reliability of the scale- combining individual questions into a scale and setting the combined scale as another variable.For example, early in the course I gave the students a short survey composed of these questions:1. I expect to do well in this course. (dowell)2. From what I know about this course, I believe that this will be an interesting course. (interest)3. This course is known to be extremely difficult.(hard)4. This course is known to be very boring.(boring)All questions have a 7-point agreement scale (1 = strongly disagree, 7 = strongly agree).The terms in the parentheses are the variable names in SPSS.I think that these four questions measure a construct called “student expectations,” and I wantto make a Likert scale with these four questions.Before I can test the reliability of the scale, I need to make sure that the items are coded properly, such that high scores measure a high level of the construct.In this case, I want high scores to measure high expectations to do well.So, I need to reverse code the responses to Question 3 and Question 4.1COMM 301: Empirical Research in CommunicationStatistics: Combining items into scales on SPSS and testing scale reliabilityThere are two ways to reverse code variables—1) recoding; 2) computing1) RecodingTransform > Recode > Into Different VariablesClick the variable you want to recode into the box. In my case, I want to recode the variable “hard,” so I click it into the box.Under Output Variable, for the Name field, type in the new variable name.In my case, I call the variable “hardr” (r at the end to indicate reverse coding).Under Output Variable, for the Label field, type in an appropriate label.In my case, I label it as “Reverse hard”.Click “Change”, to have the new information take effect.Next, click on the button “Old and New Values …”On the left side, under Old Value, for the Value field, type in 1.On the right side, under New Value, for the Value field, type in 7.Click “Add”.Repeat, under Old Value, type in 2; under New Value, type in 6.And so on, until you finished reverse coding the scale for the original question:1 becomes 72 becomes 6 …7 becomes 1.Do the same to recode “boring” into “boringr”.You need not enter the information for the old and new values again. They are saved in the program.Click “Paste”. Go to the syntax file. Highlight the appropriate section, and click ►.Go to the data editor interface to clean up any discrepancies, or add any information.2) Computing Transform > ComputeType and label the new variableIn my case, I call the variable “hardrc” to make a comparison with “hardr”Write a numeric expression as following(Number of levels + 1) – “original variable”In my case, it is “8 – hard”2COMM 301: Empirical Research in CommunicationStatistics: Combining items into scales on SPSS and testing scale reliabilityNext, I want to test the reliability of the scale. To do that, I execute this procedure:Analyze > Scale > Reliability AnalysisClick to put the relevant variables into the box.In this case, the variables are “dowell”, “interest”, “hardr”, and “boringr”.For “Model”, leave it as Alpha.Click “Paste”.Go to the syntax file. Highlight the appropriate section, and click ►.This is the output.ReliabilityR E L I A B I L I T Y A N A L Y S I S - S C A L E A L P H A)Reliability CoefficientsN of Cases = 10.0 N of Items = 4Alpha = .4105So, is this scale reliable?How can you tell?So, what can you do?3COMM 301: Empirical Research in CommunicationStatistics: Combining items into scales on SPSS and testing scale reliabilityScale constructionAssuming you have a reliable scale, the next thing to do is combine all the various variables (the different items) into one variable (the scale).For example, I want to combine “interest”, “hardr”, and “boringr” into one variable, called “perceive” (for initial perceptions).To do that, I use the following procedures:Transform > ComputeTarget Variable: perceiveNumeric Expression:(interest + hardr + boringr ) / 3Click “Paste”.Go to the syntax file. Highlight the appropriate section, and click ►.In further analysis, I will use this variable “perceive” (perceived interestedness of the class)For future reference, write “interest + hardr + boringr” in the label section. In this way, you will not forget what variables are used to compose the “perceive”