topic Re: Requesting aid in understanding how to use SAS to build a multiple logistic regression model in Statistical Procedures

Requesting aid in understanding how to use SAS to build a multiple logistic regression model

KryptonZA — Mon, 31 Aug 2020 17:15:02 GMT

Hi, so I recently got given a dataset as well as a research question, stating "which skills would be good predictors of the variable of interest."

I am a university student, who is new to SAS, so my knowledge is very limited, so excuse any confusing wording, I am currently just trying to understand everything that would be required from me.

I understand I will have to use of the variables and setup various hypothesis to evaluate whether some of the variables have correlation towards predicting the variable of interest. To my understanding the variable of interest will be the independent variable, and I will have to make use of several proc reg or proc corr statements to yield descriptive statistics which I would be able to evaluate.

How will I go about solving this problem, how will I set up hypothesis? Till thus far I could only think of "if X has a higher percent, it would lead to significant difference in the variable of interest" which seems wrong to me. How will I know which tests to run, and how to interpret the results?

Please correct me at any place where I have misunderstood or misinterpreted the situation.

Thank you in advance to anybody willing to help, I really appreciate it.

Regards.

Re: Requesting aid in understanding how to use SAS to build a multiple logistic regression model

SteveDenham — Mon, 31 Aug 2020 19:06:59 GMT

It seems that you have been thrown into the deep end of a pool, with no previous experience in swimming. That is unfortunate! My suggestion would be to let us know what your statistics comfort level is, and for you to read the documentation, especially the Getting Started section, of the PROC LOGISTIC documentation. As you go through that, perhaps you will identify things that are similar to your data and question. Also, you should probably do what you can to get a copy of Hosmer and Lemeshow's text - Applied Logistic Regression. or of Fleiss's Statistical Methods for Rates and Proportions.

On that note, I would say to stop anything you are doing right now, and go read these sources, as nothing in your presentation here looks to me like it would be helpful in addressing your research question.

SteveDenham

Re: Requesting aid in understanding how to use SAS to build a multiple logistic regression model

Reeza — Mon, 31 Aug 2020 19:25:27 GMT

Forget SAS for the time being. First figure out what you want to do in as much details as possible. You likely want a summary statistics table of some sort of your raw data for exploratory data analysis. Then you'll want bivariate analysis based on your outcome variable. Then you'll move on to modelling.

For each of these steps its a different task/process. If you're familiar with the statistical concepts but not SAS then I'd highly recommend taking the FREE SAS Statistics 1 training course. It's a two or three day course BUT you'll save more time overall with this approach.