Assignment WK 11

     

     

         

       

                                       

                                                             

                                     

             

                                                                                                       

   

                                                     

                               

                                 

                                                                   

         

                                                 

                                     

                               

                             

                                             

                       

Bivariate Categorical Tests

Bivariate Categorical Tests Program Transcript

[MUSIC PLAYING]

MATT JONES: Up to this  point, we've been focusing on statistical tests  that require metric  or  variables-­-­ that is, variables  measured at the interval or  ratio level. But there are a lot of categorical variables  that are of use to the social scientist. We're going to cover  the chi-­square test for  independence and associated measures  of effect of Cramer's  V in SPSS.

Let's go to SPSS. We can test for  the relationship between two variables  by   using the chi-­square test for  independence. Let's go ahead and test the relationship between gender  and views  on marijuana legalization.

To do this, we go to Analyze, Descriptive Statistics, and Crosstabs. Here, you will see a place to put a variable in a row and a column. I'm going to scroll down and find my  SHOULD MARIJUANA  BE MADE  LEGAL variable and enter  that into my   row, and I will scroll down to find the respondent's gender  and place that into my   column.

I'm going to go ahead and hit OK  to show you the output that we receive. And here, you will see some output that are basic  Descriptive Statistics. These are counts  of the number  of males  and the number  of females  who felt that marijuana should be either  LEGAL or  NOT LEGAL.

However, this  does  not statistically  test for  a relationship between these variables. We can request the chi-­square statistic  by, again, going back  into our   Crosstabs  box. So I perform  the same procedure of going to Descriptive Statistics, Crosstabs, and all of my  information is  still there, so I can select Statistics. You'll see that the Chi-­square statistic  comes  first, but I have to go ahead and activate that.

I'm also going to go into the section Nominal to ask  for  Phi and Cramer's  V. This   will tell me something about the strength of the relationship between the two variables. As  you know from  your  reading, the chi-­square tells  us  whether  there is  a relationship, but it doesn't tell us  anything about the strength of that relationship. Find Cramer's  V help us  with that follow-­up should we have a significant relationship with a chi-­square. Continue.

OK. So I'm going to hit Cells. Just for  ease of interpretation, I'm going to request Percentages  for  Columns.

I'll hit Continue and OK. Here, you see, I receive some Case Processing Summary. This  tells  me that there are 920 valid cases  in this  analysis. 580 cases   are missing. So out of the 1,500 cases  or  respondents  of the survey, we have

©2016 Laureate  Education, Inc. 1

     

     

                             

                                         

                                                                 

                                                                         

             

                                             

                                               

                           

                                           

                             

                                                                                       

     

 

Bivariate Categorical Tests

quite a few of them  that either  didn't answer, refused to answer, or  just left that blank.

The next piece of output is  the Crosstabulation table. You can see this  looks   similar  to the Crosstabs  I asked for  in Descriptive Statistics  with just the raw counts, but now, I also requested for  the percent within respondent's sex. This   tells  me 55% of the males  believe that marijuana should be made LEGAL and 44.6% of the males believe that marijuana should be NOT LEGAL for  a cumulative percentage of 100%. I can interpret the female column as  the same way. 41% of females  believe that marijuana should be LEGAL while 59% believe that it should be NOT LEGAL. If there was  no relationship between these two variables, we would see approximately  equal percentages.

To statistically  test for  this, we can look  to our  chi-­square statistic. Here, we see a critical value of 18.993 with an associated p-­value of 0.001. This  test is  significant at the 0.01 level and certainly  well below the common 0.05 threshold. Therefore, we can reject the null hypothesis  that there is  no relationship between the two variables  assuming that there is  some sort of relationship between gender  and position on marijuana legalization.

But once again, we don't know the strength of that relationship. We can scroll down to our  Cramer's  V correlation, which tells  us  about the strength of this   relationship. A  value of 0 indicates  no relationship whatsoever, and a value of 1.0 indicates  a very  strong, perfect relationship.

We can see here, we have a value of 0.144. So while there is  a relationship, it's important to do the follow-­up test to determine the strength of that relationship. In this  case, the relationship between these two variables, which is  statistically   significant at the 0.01 level, is  rather  weak.

[MUSIC FADING]

©2016 Laureate  Education, Inc. 2