I have the latent variables that measure different factors and base on perception, some of the I have created, some the items(reflective more than 0.32) I took from other studies. I translated into a different language from English. Now I want to use them as exogenous variables in my multiple regression. On the other hand, I have two different endogenous variables that measure the speed. I want to use the same factors to compare for both multiple regression models. My hypothesis is about the proposed measurement of speed is better than old school for a different type of sectors ( that is encoded as high tech and low tech).
For instance;
Y1: Old school measurement (continuous)
Xn: exogenous factors (latent factors with likert scales)
Cumulative : Y1= a0+a1X1+a2X2+a3X3+....anXn (n:255)
Hightechs with old school: Y2= b0+b1X1+b2X2+b3X3+....bnXn (n:81)
Low techs with old school: Y3= c0+c1X1+c2X2+c3X3+....cnXn (n:171)
N1: new measurement (continuous)
Cumulative : N1= d0+d1X1+d2X2+d3X3+....dnXn (n:255)
Hightechs with old school: N2= e0+e1X1+e2X2+e3X3+....enXn (n:81)
Low techs with old school: N3= f0+f1X1+f2X2+f3X3+....fnXn (n:171)
Comparing N1 and Y1, Y2 and N2 and, N3 and Y3
I am planning to compare same sectors with a different measurement. So, first I want to EFA because of translation; with principal axis factoring and Promax rotation. Eliminate the factors and check the CR and Cronbach alpha levels, if they are insignificant then eliminate the factor, then take the arithmetic means of the latent factors and assess the regression model. Should I compare betas or perform t- tests for coefficents? Which statistical method should I use? Should I use SmartPLS or SPSS?
Model selection may be based on some criteria like R-squared, which is called the coefficient of determination. The larger it is, the better the model.
Other criteria include Akaike information criterion (AIC), Bayesian Information criterion (BIC), Schwarz Information criterion (SIC), Hannan-Quinn (HQ), etc. The smaller they are the better the model.
Arsen,
In this case where you are looking to compare two models that have different dependent variables I don't think any of the information criterion will be useful.
As best I can follow (forgive me, I can easily confuse myself), you'd like to see which dependent variable your independent variables better explain or predict. If it's explaining you're interested in, Ette's suggestion of R2 can be useful. For predicting you might also think about Root Mean Squared Error (RMSE). Perhaps better, if you have the data to afford it, you could do some sort of cross-validation with train and test data.
There's a similar thread on Stack Exchange (https://stats.stackexchange.com/questions/178431/comparing-two-models-with-different-dependent-variables).
Oh, and as far as statistical environments go I'm keen to R or Python. Highly flexible, extremely powerful, steep learning curve.
Best,
Caleb
Hi Arsen
1. What is the method of estimation used in your regression models?
2. The method should provides the measures: R square , R square change, values of Significance F, the residual and the significance values of the parameters of all models.
3. Based on above measures, you can differentiate the models.
Regards,
Zuhair
Sorry, please refer to:
1. Landau, S. and Everitt, B. S.(2004). A Handbook of Statistical analyses using SPSS.
2.Howitt, D. and Cramer, D.(2008). Introduction to SPSS.
Regards,
Zuhair
Thank you all for your contribution, what if i use t test to compare coefficients to find is there any significant differences?
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA