Curvilinear Regression - SPSS (part 2)
Вставка
- Опубліковано 14 жов 2024
- I perform a curvilinear regression analysis in SPSS. Specifically, I test a quadratic effect (one bend in the regression line) using a hierarchical multiple regression approach. I point out the key to the analysis, which is the F change value associated with the squared independent variable. I discuss the beta weights and how they are not particularly interpretable. I also discuss multicolinearity and why it is not a problem in the nonlinear regression case. I also show how to do the nonlinear analysis using a second approach in SPSS which gives more useful scatter plots in the nonlinear regression case.
Thank you so much. Now what if I need a logarithmic regression?
Hi i have 1 question. When i want to use the best curve form my graphs i go to analyze/regression/curve estimation and then i choose all models. Then in output i get R square for all curves. My question is how i know wich curve is the best to use? The one with the biggest R square?
thanks for the video!
quick question: my predictor variable is market concentration (0-100%) - if i square it my squared concentration value is lower than my initial value. e.g. concentration of 0,2 (20%) becomes concentration squared = 0,04 (4%). so how do i get my second value? should i take the sq root of the first term and end up with sqrt(0,2) and 0,2 for my values?
thanks!
+Justus Tomczak Instead of using proportions (e.g., .20), why not use percentages (e.g., 20%)?
+how2stats
sure but that gives me values >100% - wouldn't that be unrealistic for a measure of market concentration?
pardon my ignorance, im a dummy in statistics, But the F value of model 1 is 49.075 and the F value of model 2 is 37.037 correct? Then why is the change in F value 22.017 and not just a substraction of 49.075 minus 37.037?
thank you very much. i have one question here. do we need to standardise IQ between the multiplication?
No, you do not need to standardise the data.
Can you share the data used in this example?