Entering edit mode

19 months ago

ohtang7
▴
40

Hello,

I am now working with a Linear Model (and also GLM) with my data in R.

I made my model, but one thing is very awkward to me.

Below is my data format.

And my model summary is as below.

In the variable part in the summary, you can see the categorical word 'None, Non-Central, urban, Yes' are behind to the words of variables.

I didn't see any models of the examples I've seen in the study showed this kind of result.

Is there anything I missed some option in the model script ?

What' the reason and How can I fix this phenomenon in the result of model ?

Thank you.

This is just how R shows results for coefficients following the automatic construction of contrasts from a design.

For instance,

`"HomelessNone"`

means that for the factor`"Homeless"`

you are testing the difference between the level`"None"`

and the reference level, which I take to be`"Exist"`

for your data.So what this actually means is that the estimate for the coefficient shows the change in

`shannon_entropy`

when`"Homeless"`

is`"None"`

compared to`"Exist"`

, while taking all the other covariates into account. Same goes for other variables: Non-Central vs Central, Yes vs No, etc.This is when you have only two variables, where the reference is taken to be the first level of the factor, and contrasts are built for the other level. If you had other levels you would have e.g.

`"subway_stationYes"`

(vs No) and`"subway_stationMaybe"`

(again, vs No) as different coefficients.If you want to change your reference level, and assuming that your

`alpha$Homeless`

and other variables are of class`factor`

, you can doIn which case the summary of the glm will show the coefficient name as

`HomelessExist`

, and you know you have to interpret the coefficient as Exist vs None.Thank you for your reply. Your explains really helped me a lot. ^^