Nagelkerke R2; P values

[S Sam]

Hi PRSice users, In the bar plot showing the model fit (R squared on Y axis)  at broad P value thresholds (on X-axis), could anyone please help me understand how to interpret the 'P values' that are printed above each vertical bar? Thanks

[Shing Wan Choi]

Those p-values are the significance of the regression for that particular threshold. 

[Nikman Adli Nor Hashim]

Let say, your base association results data (ie. GWAS) is for autism phenotype. And your target (data you want to test on) genetic data contains depression status as the phenotype. 

As I understand it, the best threshold for depression status to fit with (or can be predicted by) the autism PRS is around 0.44.

This broad threshold gives the most significant genetic relationship (ie. the lowest P-value) between the mean value of the autism PRS, and depression status (via regression).

If you want to have a more specific threshold value, the hi-res PRS plot will show you where it lies within the bar plot.

Sometimes, you can get all/many 'red' bars' with very low P-values for those thresholds... In that case I suggest looking at both your base and target data, whether there are related/same individuals within those two groups of data. The crucial thing is not to have the same individuals/related individuals in them because the it would cause overfitting (This issue can also be used to test your own cohort's PRS, by using it on itself. It should overfit in most cases).

Please correct me if I'm wrong, and I hope this helps. :)