Independent Samples (binomial regression)

I have a set of destroyed and not destroyed homes from a Wildland-Urban interface fire. I have found through a chi-square test for dependence that many of the surviving homes (not destroyed) are dependent on fire fighters performing water suppression on the adjacent destroyed homes. In other words, the fire fighters stopped fire spread from burning homes to adjacent not burning homes.

Other factors might have contributed to the fire fighters ability to stop the fire and I am thinking about checking those other factors, initially through binomial regression.

Because of this dependence I have found between destroyed and not destroyed homes, does this sample violate the assumption of independent samples as many not destroyed homes survival were dependent on fire fighter actions on adjacent destroyed homes?

Thanks for any help folks can provide!!!


Active Member
What exactly do you try to check, what is your model?
You may say that the quality of firefighters is one IV, and also the water pressure is IV.

So the DV is binary? Destroy or Not destroy? (why not logistic regression?)

In any regression, there is a dependent variable and independent variables...
Independency should not exist between the observations.

I can see that there may be some dependency between the observations, "herd immunity ..." that you should try to reduce. unsafe home in a safe neighborhood.

But please show the model you build.
Thanks for the reply!!

I do not have a model yet and I am mainly asking about statistical tests that have an assumption of independent samples and if my population violates the assumption of independent samples (I understand there are dependent and independent variables).

So, imagine there are 100 homes affected by a Wildland fire. They are organized in groups of two adjacent homes, so fifty groups, lets say. Let’s say all groups had at least one destroyed home, some had two destroyed homes. Let’s say all groups that had one destroyed home had water suppression on that destroyed home, and all groups that had two destroyed homes had no water suppression on any of the homes. Let’s say we determined unequivocally that surviving homes where dependent on water suppression on the adjacent destroyed home.

If the above is the case, then are the destroyed and surviving homes not independent samples and this population violates the assumption of independent samples for statistical tests that have that assumption.

Obviously sometimes it is easier than others to determine if samples are independent of each other and I am trying to determine what is the case for this hypothetical sample.



Fortran must die
Assumptions about independence, except for time series or repeated measure designs, generally involve how you gather your data. There is no way I know of in cross sectional designs to test if the assumption of independence is met.