Degrees of freedom for a chi-square goodness-of-fit test with estimated parameters and merged cells

A frequency data is classified in 10 classes and normal $N(\mu,\sigma^2)$ distribution is fitted to data after estimating the parameters. To use $\chi^2$ goodness of fit test, two classes are combined into one class. Hence the degrees of freedom associated with $\chi^2$ test are: