0

I did 2 linear regression analyses. White woman (WW) N1= 50 Black woman (BW) N2= 49

Correlation btw WW's V1 & V2 P = 0.001 Correlation btw BW's V1 & V2 P = 0.013

How do i know if the 2 p values are difffernet from each other in order to say that V1 & V2 of WW is more associated than that of BM for example? I just want to know how to compare p values when the sample size is different!

Thank you!

tsumaranaina
  • 191
  • 7

1 Answers1

1

You should not compare models using p-values.

What you really want to know is: Is the correlation for White women different than for Black women?

You can use this by running one linear regression analysis. Let's call your outcome $y$, your independent variable $x$, and a variable that says what race they are (coded 1 for White and 0 for Black) is $z$. You can do an interaction in one linear regression model:

$\hat{y}_i = \beta_0 + \beta_1x_i + \beta_2z_i + \beta_3x_iz_i$

where the $\beta$s are regression coefficients. What you are interested in is the interaction, or $\beta_3$.

Mark White
  • 8,712
  • 4
  • 23
  • 61
  • Thank you very much. So this means that if i am checking the salary of WW with eating salad, and that of BW as well. Eating salad would be y. salary would be x and race would be z, right? Is there a way to do this in SPSS or do i have to use R? *Also, sorry i cant upvote because i don't have enough reputation.. – tsumaranaina Jan 04 '18 at 04:20
  • There are a lot of guides online on how to do it in SPSS (https://stats.idre.ucla.edu/spss/webbooks/reg/chapter7/regression-with-spsschapter-7-categorical-and-continuous-predictors-and-interactions/). You basically create a new variable that is `salary * race`, and include this variable (along with salary and race) into the model. – Mark White Jan 04 '18 at 05:34