The estimated parameter in an AR (1) with just one dependent variable is 0.92. I have checked the residuals for heteroscedasticity and both the Breusch-Pagan test and the White test confirm the presence of heteroscedasticity. Does this affect the reliability of the estimated parameter? Can I still confirm that the time series is AR (1) with an Autoregressive coefficient of 0.92?
If you used OLS or conditional least squares, your estimate is still consistent although biased (but bias is hard to avoid for autoregressive models).
More importantly, it is inefficient due to the heteroskedasticity. But that does not necessarily mean you can do better (that a feasible efficient estimator exists).
In any case, you cannot "confirm" the estimate is exactly 0.92, but you could obtain a confidence interval around your estimate (using some heteroskedasticity-robust variance estimator) and phrase your conclusion in probabilistic terms.