Chi Squared Tests

SPSS: Chi Squared Tests

So, let’s say you count some (1851) plain M&Ms, and here’s what you find. Also, notice that the information from the M&M/Mars Corp. is that 30% of plain M&Ms are brown, 20% each yellow and red, and 10% each of blue, orange, and green. In Excel, it is not too hard to type in the formulas for the ChiSquared tests, but you can also do it in SPSS. First, we need to enter our data into SPSS. Notice that I labeled the columns, and arbitrarily assigned numbers to the colors (I could’ve used any numbers, but going in order makes the most sense). I also typed in the value labels on the variable view screen (so, 1=brown, 2=yellow, etc. on any printouts). However, SPSS typically thinks of each row as an individual. Therefore, if we ran our analysis now, it would think we had 6 M&Ms, not 1851. To run our test the “normal” way, we would need to enter 1851 rows of data! To get around this, we use a command called Weight Cases. Weight Cases This feature is a way of telling SPSS to pretend that a single row of data is, in fact, a whole bunch of rows. In our first row, we want to tell SPSS that we have 519 brown M&Ms. So, look for the menu item: Data: Weight Cases When you do, you get this pop-up window:

The Default setting is “Do not weight cases”

Putting a Variable in here means, “Consider each row to really mean this many rows.”

Now, SPSS will pretend that the variable “color” has 519 rows with a “1,” 364 with a 2, etc. This will be in the bottom right corner of your screen, if you did it right. You can then do the Chi-Square (or any) test like you would’ve before. Except now, the six rows will act like it is 1851 rows.

Page 1 of 4

SPSS: Chi Squared Tests

Chi-Squared Test of Goodness of Fit So, our Null Hypothesis will be that the probabilities given to us by the corporation are correct: H0: Plain M&Ms are 30% brown, 20% each yellow and red, and 10% each blue, orange, and green. H1: Plain M&Ms have a different distribution than that. This...