Once you have determined that you have violated the assumption of homoskedasticity of prediction errors in the context of OLS regression, then you may need to come up with an alternative strategy for estimating regression parameters - and most importantly, the standard errors. Weighted least squares (WLS) is one such option. This video provides a brief illustration of steps for carrying out weighted least squares (WLS) regression in SPSS.

A copy of the data can be found here to download and follow along:
https://drive.google.com/open?id=1Nyx...