Inferential Data Analysis: Techniques and Interpretation

Task:

After you have done the discussion of your descriptive data analysis, your next step is to conduct inferential data analysis. Now you are moving from just describing the individual variables of interest to investigating whether there are statistically significant relationships among your variables of interest. These inference tests will help you answer your main research question.  

Your inferential data analysis should include more than one inferential test. This could just be an ANOVA with an examination of the overall F-test and discussion of the post-hoc tests. It could also be a chi-square and a logistic regression. Any of the tests of inference discussed in Chapters 7 and 9 of your textbook are fair game for you to use.

For your second draft paper you need to submit the following:

1. Short description of your research question

2. List of the SAS analyses you did to provide inferential statistics

3. Discussion of the results of those analyses

4. Your code

5. Your results

Make sure your second draft is in APA format. 

HOW TO DO IT IN SAS

(Paired t-test, independent t-tests, chi-square and odds ratio were discussed previously, these are just additional analyses you might want to do.)

QUICK TIP: Remember, you can use a BY statement (make sure to do PROC SORT first) in any of these SAS procedures if you want to stratify your analysis by a certain categorical variable, like gender. But only do this, if it is relevant to your research question. 

HOW TO DO IT IN SAS

(Paired t-test, independent t-tests, chi-square and odds ratio were discussed previously, these are just additional analyses you might want to do.)

1. Find a correlation between two or more variables

PROC CORR DATA = datasetname;

VAR variable1 variable2 variable3;

RUN;

You can list as many variables as you want. SAS will give you a matrix of correlations.

2. Compute an ANOVA with one independent variable 

PROC GLM DATA = datasetname;

CLASS categorical variable;

MODEL dependent variable = independent variable;

MEANS variable / TUKEY;

RUN;

Note:

§ You use the CLASS statement whenever you want to include a categorical variable in your ANOVA or regression analysis.

§ Make sure the variable that you place next to the MEANS statement is a numeric variable, since means are appropriate measures for numeric variables.

3. Compute an ANOVA with two independent variables and an interaction effect 

PROC GLM DATA = datasetname;

CLASS variable1 variable2;

MODEL dependent = variable1 variable2 variable1*variable2;

MEANS variable1 variable2 / TUKEY;

RUN;

Note:

§ Any categorical variable in your analysis must be included in the CLASS statement.

§ Whenever you have an interaction term in your model, you must make sure to include each of the individual variables that make up the interaction term. For example, if I have an interaction gender*height, then I must make sure to include gender and height in my actual model.

4. Compute a linear regression with multiple variables

PROC GLM DATA = datasetname;

MODEL dependent = variable1 variable2 variable3;

RUN;

5. Compute a logistic regression with multiple variables

PROC LOGISTIC DATA = datasetname;

CLASS variable;

MODEL dependent = variable1 variable2;

RUN;

Note: In logistic regression the CLASS statement is optional and only needs to be used if you have categorical independent variables.