I have a model that links the heating of a component compared to the frequency applied to it.
The model is $\text{Heat} = \alpha \cdot \text{Frequency} + \beta$.
Based on the data, I found $\hat{\alpha}=-0.06332$ and $\hat{\beta} = 0.35139$. The observed variance is $S^2=0.0572$. There are $n=8$ sample data.
Now I already have determined that since $T_\text{obs} = 2.1687 > t_{0.95;6}= 1.943$, the frequency has an impact on the heating.
Next step in the exercise is to find if a new model with a second parameter (the length) is better or no. The exercise states that the new model has an $SSR_2 = 0.05$.
I can compute the first model's SSR with $SSR_1 = S^2\cdot(n-2) = 0.0572\cdot6=0.3432$
Here is the part I struggle with. The formula I have is $$F=\frac{(SSR_0-SSR)/R}{SSR/(n-p)}$$ And I would compare it to the table's value for $F_{0.95}(6,5)=4.95$ (because I have 8 data and 2 parameters for the first model, and three for the second model).
I think the parameters are $R=1$ (because there is one parameter difference) and $(n-p)=5$ (because 8 data and 3 parameters in the new model).
I don't know what to put in $SSR$ and $SSR_0$
$H_0: \text{Reduced model (freq. only)}$ vs. $H_1: \text{Full model (incl. length)}$.
Hence, $SSR_{0}$ refers to the residuals sum of squares of the model under $H_0$ and $SSR$ refers to the full model. Now, recall that $F$ is a ratio of Chi squared r.vs, hence, the $R$ denotes the number of Restrictions. Namely, you restricted in the reduced model the coefficient of length to be $0$, hence $R = 1$. And the $n-p$ in the denominator belongs to the full model, i.e., $n - p = 8 - 3$. Namely, you have to compare the calculated $F$ to $F_{(1, 5)}(0.95$).