This site uses different types of cookies, including analytics and functional cookies (its own and from other sites). To change your cookie settings or find out more, click here. If you continue browsing our website, you accept these cookies.
I've gotten this error previously using Logistic Regression and I believe (but am not certain) that what solved it was making sure that I excluded any columns that have only one constant value (eg all the values are 1's) from the list of predictor variables.
I agree that having your exact workbook might allow us to help further if there is something further.
It is important to remember/know that all of the predictive tools are actually Macros (.yxmc). That changes a few things during troubleshooting. It turns out that a macro is just a workflow wrapped up behind a tool icon/interface. It has its own tools/messages/errors/warnings but by default, only errors float up to the workflow/module where you are using the predictive tool (macro). When you get errors it can be helpful to see ALL the messages (not just errors). I find that the message right before or right after the error often tells me what is wrong.
Go to your Workflow Properties: Runtime Tab. Select the check box to 'Show All Macro Messages' and rerun your workflow. Check out the new messages you'll see and let us know if that helps!
Fit Stats is an object in the final report (what comes out of the R output from the Logistic Regression tool). For some reason it isn't successfully created due to a data issue (perhaps a long variable name, missing data, special characters, or a number of other things R does not like) that the macro messages might tell us more about!
Thank you for the solution. I have solved the problems yesterday. But you answer also helpful, cos I don't know there is the "Show All the message" before.
I still have a small question, it is about the how the model running.
In Predictive - Logistic Regression, I have two kinds of variables.
One of them are continuous variable, such as X1= [1,2,4,5,7,8,13,,,,,100] , X2= [0.11, 0, 23, 1.24, 2.33, ... 9,87, 9.96]
The other variables are differents levels, such as X3= [10%-20%, 10%-20%, 10%-20% , 20%-30%, ...., 80%-90%, 80%-90%,, 90%-100%], all the data is belong to one of the category.
For example, there are 10,000 rows data, so there are 10,000 X1, 10,000 X2, 10,000 X3. If I set X1, X2, X3 as the predictor variables, it will be running extremely slow, and finally I got nothing, just the error"Error: Logistic Regression (2): Logistic Regression: Error: cannot allocate vector of size 3.2 Gb". But if I changed X1 into several categories, like X1= [1-10, 1-10, 11-20, ....91-100 ], X2= [0-1, 0-1, 1-2, 1-2, ...9-10], then the model works, the error solved.
So here is the problem,
1. What is different to use the continuous variable directly VS use the categorised variables?
2. Why is it extremely slow to use the continuous variable and result in the memory problem?
Is it possible that Alteryx thinks that your continuous variable is actually a string and not a float or double. If the tool is trying to evaluate your number as a string (it would try to create a factor for every single value, I think) and it would likely error out.