The possibility of a correlation between the radio (cm)- and $\gamma$-ray luminosity of variable AGN seen by EGRET is investigated. We performed Monte-Carlo simulations of typical data sets and applied different correlation techniques (partial correlation analysis, $\chi^2$-test applied on flux-flux relations) in view of a truncation bias caused by sensitivity limits of the surveys. For K-corrected flux densities, we find that with the least squares method only a linear correlation can be recovered. Partial correlation analysis on the other side provides a robust tool to detect correlations even in flux-limited samples if intrinsic scatter does not exceed $\sim 40$ \% of the original $\gamma$-ray luminosity. The analysis presented in this paper takes into account redshift bias and truncation effects simultaneously which was never considered in earlier papers. Applying this analysis to simultaneously observed radio- and $\gamma$-ray data, no correlation is found. However, an artificial correlation appears when using the mean flux. This is probably due to the reduction of the dynamical range in the flux-flux relation. Furthermore, we show that comparing the emission in both spectral bands at a high activity state leads to no convincing correlation. In conclusion, we can not confirm a correlation between radio and $\gamma$-ray luminosities of AGN which is claimed in previous works.