Question

Suppose Heat Power developed a regression model relating heating average annual pay to the percentage of households using natural gas as heating type. Below is the plot of the corresponding residuals versus the predicted values. Versus File free Rated Value What does the residual plot suggest? The spread of the residuals decreases as the fitted value increases. The plot of residuals shows no unusual pattern. The Equal Variance assumption is satisfied. The spread of the residuals remains unchanged as the fitted value increases. The Equal Spread condition is satisfied. The spread of the residuals increases as the fitted value increases. The equal spread condition is not satisfied. The plot of residuals has no particular shape. So, the Equal Spread condition is definitely satisfied. Considering your answer to the previous question, which of the following data transformations would be appropriate as a good start point for the response variable (average annual pay) to make it more symmetric and/or simplify its relationship with other variables? Square the data values. Only extrapolation would help. Use a logarithmic transformation. Use a geometric transformation. Cube the data values.

Answer 1

DATE * be n so see is 4th i e The given plot given plot is such that the residuals can accommodated inside a double bow then such pattern indicates the variance or errors is not constant but y in proportion betwee and from plot we also that the spread of the residuals increases as the fitted value increases then the equal spread condition not satisfied option is correct The spread of the residual increases as the litted value increases. The equal spread Condition is not satisfied residual can be accommodated value shows a double bow pattern then y proportion between and I y then may have a Binomial distribution. The variance of binomial distribution proportion hear 0.5 greater as compared to hear so, the assumed relationship between are non-linear. we must have change in extrapolation. The following table shows the transformation for given conditions. and option is correct. vs fitted O The is zero Or y and x's so as a CS Scanried with CamScanner

Transformation Relation of oto E(y) o constant O'E() O.E(v)[1-E()] Ox[E(1) [E(y) y = y (no transformation) yo = (Poisson data) y* = sin(V) (Binomial proportion Os y, 51) y =In(y) y* = 11 ty 1 o «[E(y)] V CS Scanned with CamScanner