Assumptions

Hey guys,

Here are some responses to questions we've received about the assumptions you should be testing for:

• All relevant IVs are included in the model: this is based on theory, so you will not have a statistical output to "prove" that this assumption has not been violated. However, you should be able to explain that the relevant IVs (that pertain to your hypothesis and what you are testing/researching) are included and measured well.

• No measurement errors in the IV's: Again, this is done prior to the data collection. You want to make sure that the measures used to assess the IVs are reliable and valid.

• NOTE: Dr. Kim has not asked you to write up anything about the two assumptions above on your final project. However, they are important to consider in all future research.

• The relationships between IVs and DV are correctly specified: See the comments provided by David on Diana's questions about the final project (Thurs, June 2). You will be able to see a clear violation if one exists for this project.

• Independence of errors: check this with the residual and predicted plot….and Dr. Kim said “Basically, you don't want to see any specific pattern among the residuals that should be randomly distributed around the zero line.”

• Homoscedasticity: Create a scatterplot with the ZPRED and ZRESID and see if there is a pattern of systematic error. See the notes and powerpoints.