Hello Atikhom,

Which particular statistic in the output was omitted? Which version of spss are you using?

Are you willing to post your data and proposed model, so that others could attempt to recreate the condition?

Some obvious points to consider about your data set whenever unexpected results such as the one you report occur:

1. Are your data for any continuous variables suitably scaled? (so that the leading significant digit is not many orders of magnitude away from the decimal point)

2. For your categorical variables, do you have cases in each possible cell/level? (check frequencies for each such variable)

3. Do you have any instances of perfect relationships among categorical variables (perhaps due to empty cells)? (check cross-tabulations for variable pairs and sets)

4. Is one of the IVs redundant with another variable in the data set?

5. Do you have missing data (and are attempting some sort of imputation process)?

That may not cover the waterfront, but at least it may give you some ideas when checking your data.

Good luck with your work.

Dear Prof. David Morse

Thank you for your detailed responses. I really appreciate them. After I read your answers, I just realized that all the individual verbs under observation are assigned as two predictor categories: semantic and pragmatic aspects. (This is my mistake since I am not good at statistics.) This means that these two observations (i.e. semantic and pragmatic) are dependent, thus violating the assumptions of logistic regression. This is evidenced by the fact that when I deleted either of the categories from the model, the issue no longer persists; the model is satisfactory.

Previous literature indicates that semantic and pragmatic aspects predict the choices of the outcome variables. Since this is the focus of my investigation and literature suggests that semantic and pragmatic aspects have effects on the outcome variables, what should I do? Should I build a different model since these two variables are not independent of each other? Thank you for your help.

Best regards

Atikhom

Hello again Atikhom,

I'm glad you found the apparent problem.

However, I'm not fully clear on what you're trying to do. Could you give a concrete example of the chosen IVs and target DV? Then I and/or others could perhaps offer some more focused advice.

Good luck with your work.

Based on this issue, the predictor variables (IV) are two categorizes of reporting verbs: meaning group (has four values - argue, find, show, think) and pragmatic group (has four values - research, cognition, discourse, and evaluation). The outcome variables (DV) are four choices of tenses, namely the present simple, the present perfect, the past simple, and the modal verbs.

In addition to other predictor variables, I'd like to observe the effects of these predictor variables on the choices of tenses. The issue is that when I entered the two predictor variables simulteanously as the independent variables and the choices of tenses as depedent variables, there was the message "Floating point overflow occurred while computing this statistic. Its value is therefore set to system missing." I'm not sure how I should solve this problem while I can still keep the two predictor variables.

Best regards

Hello again Atikhom,

It sounds to me as if all three variables are nominal and each has four categories/bins. Based on this, I don't think that multinomial regression is likely to be your best option.

Instead, you might consider a log-linear model, or chi-square independence method. Of these, the log-linear model is better suited to interpretations along the lines of "main effects" and "interaction" terms.

Good luck with your work.