Check my residual graph (please!)

#1


IVs are precentage of students eligible for fsm, percentage of non white students, total number students in each school, and whether not school classified as 'failing'.

The x axis is the dependent variable, and the y is standardised residuals. Is this correct? The books I've looked in say to put predicted value on x axis and residuals on y, but there isn't an option in SPSS. Or is that what my graph actually is (confused)?

Does this graph mean the regression is null? It's population level data if thet matters

Any help well appreciated.
 
#2
For the benefit of anyone else who needs to know how to do this, I ended up saving new variables (standardised residuals and standardised predicted values) in the regressin options box (click 'Saves...') then producing a scatterplot with these two variables. The result is:



To anyone in the know: Is this acceptable? :confused:
 

mp83

TS Contributor
#3
Generally, your standardized prediction is not bad. You have things out of normality to concern you. But, look at the red thing I;ve cycled through, you have predictions under your fitted values and it;s kind of systematic nature. Probably it's not statistically significant,but interesting nevertheless.

I'm in a hurry, but I'll be back if something comes in mind
 
#4
Hi mp83, thanks for your input.

Yes that line is curious. I've been trying to work out what would cause it. I figured it might be schools that have either 100% or 0% free school meal eligibility (as this is an indicator or poverty) are special cases (say eithet independent or on special measures). So I removed these from the analysis but nothing changed. Then I removed the whole free school variable, and still there was that anomaly. Whatever I use to predict 'students passing english exam', the anomaly is there. But the distribution of residuals from the model is normal in a histogram, and the distribution of the dv itself is normal. So what gives? Does anyone know what this type of error is called? I'm guessing its a type of heteroskedasticity?