i made different models . in first I took a dependent variable and four independent variables . in second model I took different dependent variable and similar independent variable like wise I made four models but when I ran binary logistic regression I found similar p values in all models despite of different dependent variables could this happen or am I making any mistake actually I code the dependent variable as 0 and 1 in all models and independent variables in all models are same as bmi, whr, age and % body fat then p values in binary logistic regression become similar I am confused here rather the dependent variables are also linked with each other but in original test results they have different values
0
Add a comment
|
2 Answers
$\begingroup$
$\endgroup$
1
"Similar" p values can certainly happen, especially if the dependent variables are related to each other.
However, without seeing your code it's not possible to say for sure what you did or whether it was a mistake.
E.g. suppose one DV was "Voted for McCain" and another was "Voted for Romney" and another was "Voted for Obama in 2008". Those would give very similar p values.
answered Mar 9, 2016 at 20:09
-
$\begingroup$ as I have coded the dependent variables in all models as 0 and 1 , independent variables are same in all models . the p values of independent variables in all four models appear same. I am giving the model here model 1: fpg (DV)= BMI+WHR+AGE+%FATS $\endgroup$– syedaMar 10, 2016 at 17:34
$\begingroup$
$\endgroup$
4
I don't think you're necessarily making a mistake. One possibility is that the different sets of independent variables that you're choosing are heavily correlated to each other (across different sets).
-
$\begingroup$ independent variables are same in all models . the p values of independent variables in all four models appear same. I am giving the model here model 1: FPG (DV)= BMI+WHR+AGE+%FATS MODEL 2: SENSITIVITY (DV)= BMI+WHR+AGE+%FATS P_VALUES FOR BOTH MODELS IS .OOOO , .002 , .000 , .003 $\endgroup$– syedaMar 10, 2016 at 17:46
-
$\begingroup$ I am confused here why it is producing the same p values for two different dependent variables if I keep gender in place of fpg or sensitivity the p value again become same , what I am concluding from this is that it is not considering the values of fpg which I collected in my data so what will the benefit of collecting a huge range of parametric values if we donot have to put this result evaluation and what will the benefit of making different models $\endgroup$– syedaMar 10, 2016 at 17:47
-
$\begingroup$ if I have to code every dependent variable as 0 and 1 software is considering every value as o and 1 whether this is not the case suggest me what is the mistake or I am running any wrong test $\endgroup$– syedaMar 10, 2016 at 17:47
-
$\begingroup$ Can you run a cross-tab on each pair of dependent variables and see how different they are from each other? I suspect if (perhaps due to how you are coding those binary flags) that those dependent variables are very similar -- or even identical. $\endgroup$– VishalMar 10, 2016 at 17:53