Dear Colleagues... I am using Partial Least Squares (PLS) to predict student satisfaction (7 point Likert scale (1: strongly disagree to 7: strongly agree)), one of the independent factors in the model is gender diversity (1: male, 2: female). Firstly, I used Pearson correlation to find if there is a correlation between independent factors (e.g. gender) and dependent factor (e.g. satisfaction). The results showed that there is no significant correlation between gender and satisfaction.
My question is: Should I enter this factor (gender) to the model as other independent factors even if it is not significantly correlated with the dependent factor or should gender be excluded from the model.
The number of independent factors is 6.
The number of observations is 70.
Many thanks
Dear Theo,
Thanks very much for this valuable explanation.
What I have understood is that I have to enter all variables to the PLS model even if there is no significant correlation between them. In my case, I found as you mentioned that two factors are uncorrelated, but in the PLS model the independent factor had direct significant influence on the dependent factor. Therefore, I thought that there is something wrong in my analysis. Hence, I am going to repeat all analysis to enter the excluded independent factors in my previous one.
Thanks again for your great answer.
Hope you all the best, Ahmed
Dear Theo,
Thanks again for your comments.
I know that larger sample is better to get more accurate results, but this is my sample. Therefore, I will use multiple regressions instead of PLS. I have found that Harris (1985) suggests that the number of participants in this test should exceed the number of predictors by at least 50 (i.e., total number of participants = the number of predictor variables + 50). Is it feasible to follow this suggestion in my case.
Best wishes,
Ahmed
Dear Teo,
I cannot express my thanks for your quick response and valuable comments. I conducted multiple regressions and the results now more clear than before.
Thanks very much and wish you all the best.
Ahmed
Recode your gender variable as "1" and "0" otherwise. Keep in mind that we don't have two variables in isolation and it is very important to test the relation between two variables keeping the effect of other variables constant. Moreover, in response to your question "Should independent factors significantly correlated with independent factor to enter them.. " you should avoid having multicollinearity
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling