I'm currently working on my thesis where I performed a multiple linear regression. The analysis is basically about the impact of projects. Among other independent variables I have the dummy variable team composition which is divided into functional(0) and cross-functional(1).

Independent Variable                       N
Team composition                          77    
     Functional teams (0)                 70
     Cross-Functional teams (1)            7

As can be seen the number of observations (N) is quite different between those two. From the regression analysis it turned out that the team composition is not statistically significant which was contrary to what I actually expected.

Now I'm trying to figure out what could be the reasoning behind such a result. My question is on whether the huge difference in the number of observations between functional and cross-functional could be one of the influencing factors for the missing statistical significance in the regression analysis??



It's not the difference that's the problem, it's the small number in the smaller group. If you had 700 and 70, then much smaller effect sizes would be significant. If you had 38 and 38, smaller effect sizes. I tried this

IV1 <- c(rep('A', 7), rep('B', 70))
IV2 <- c(rep('A', 38), rep('B', 38))
IV3 <- c(rep('A', 70), rep('B', 700))

DV1 <- c(rnorm(7, 0, 2), rnorm(70, 1, 2))
DV2 <- c(rnorm(38, 0, 2), rnorm(38, 1, 2))
DV3 <- c(rnorm(70, 0, 2), rnorm(700, 1, 2))

lm1 <- lm(DV1~IV1)
lm2 <- lm(DV2~IV2)
lm3 <- lm(DV3~IV3)

lm1 shows p = .47, lm2 p = .04, lm3 p = .001


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.