In this guide, we'll walk through how to: X Creates bivariate contingency tables X Distinguishes the row and column percent X Significance Testing a connection in a cross table X Brings up a measure as the strength adam lanza X Controls adam lanza the relationship for an additional variable in a trivariat cross table X Constructs an effect parameter table for the trivariata table
One of the simplest but still useful statistical analysis techniques is to make so-called cross-tables. In them you can see if there is any correlation between the variables that can be nominal, ordinal or interval scale level. A cross table simply adam lanza means that one expects percent. Based on a cross-table can, for example, examine the young vote for parties on the left side to a greater extent than the old ones.
In my example figured I use data from the US General Social Survey, which is a large survey among a representative sample of Americans. The data is free and easy to download, and includes lots of fun variables. In the example, we will based on the 2010 survey to examine whether it is so that those who voted for Barack Obama for President in 2008 in higher usträckning know that man evolved from animals adam lanza compared to those who voted for John McCain.
How to create a bivariate cross table The independent variable here is named "PRES08", and has three valid response options: adam lanza vote for Obama, voted for McCain, or voted for someone else. I enter in the column "Missing" to all who voted for someone other than Obama than McCain will be deleted from the analysis, because I only want to compare the two candidates' supporters.
Cross table is created by entering the "Analyze-> Descriptive Statistics-> Crosstabs". In the dialog box, you can select the variable you want in the table's columns and what you want in the table's rows. You may do as you wish, but it is important when then expects percent to remember where to put the independent variable. Therefore, I have the habit of always adding the independent variable in rows, and the dependence of the columns, as in Figure 1. This reduces the risk of errors.
Then press the button "Cells". We are now in the "Percentages" option to get radprocent, column percent or total percent. Total percentage simply shows how many of all respondents found in each box in the table, which rarely is particularly interesting. Row and column percentages are very interesting. We click in both.
Now press the "Continue" and then "OK". We can now print a table that looks like in Figure 3. We can now see that 204 people voted for Obama and believe that man evolved from animals, while 101 people voted for Obama and do not believe that man evolved adam lanza from animals. 73 people voted for McCain and believe in evolution, while 124 people voted for McCain and do not believe in evolution. The numbers are not so interesting - they vary of course, depending adam lanza on how many people asked in the survey. There are interrelationships that are interesting, and to understand them we must look at the percentage. adam lanza
The difference between the row and column percent in each box contained two percentages. The top percentage in each box labeled "% within VOTE OBAMA OR MCCAIN" and the bottom labeled adam lanza "% within SCI KNOWLEDGE HUMAN BEINGS DEVELOPED FROM ANIMALS". The upper percentages add up to 100% in each row, and is therefore called for radprocent. The lower percentages add up to 100% in each column, and is therefore called for column percent.
With radprocenten we can answer the question "Do those who voted for Obama in evolution to a greater extent than those who voted for McCain?". We can see in the table that 66.9% of those who voted for Obama believe in evolution, while only 37.1% of those who voted for McCain does. Thus it seems to be a connection here.
Column percentages, we can instead answer the question "Voted those who believe in evolution at Obama more often than those who do not believe in evolution?", Which is an entirely different matter. We can see in the table that among those who believe in evolution as voted 73.6% for Obama, while only 44.9% of those who do not believe in evolution did it. It may be interesting, but you can not use column percentages to answer the first question! If you do that, it risks being wrong, which is a simple example can illustrate.
Say that you want to investigate whether intoxication increases the risk of drowning adam lanza when driving the boat, and therefore examine how many of those who died when they drove the boat that was drunk, and how many were sober. It is not unlikely that we would come to the conclusion that just as many were drunk as sober. If we then conclude that it is safe to be drunk while driving the boat, we are completely wrong. Why? Well, it was probably many more who drove the boat sober than who drove a boat full. The right question to ask instead adam lanza is "How adam lanza many of those who were drunk when they drove the boat drowned, and how many of those huh
