Statistics Questions with CORRECT Answers (Verified Update)
Q1: why is the mean the best estimator (in most cases) for a normal distribution?
Answer: more efficient and consistent
Q2: when using the mean and SSE, always--?
Answer: check for outliers
Q3: 1 --> n - p - how do you convert the simple model for the MSE into the complex model?
Answer: n
Q4: what is the question we consider when we're deciding whether we can reject the null
hypothesis (β0 = B0)?
Answer: is Model A better enough than Model C
Q5: analysis of variance (ANOVA) table
Answer: summary of the results of analysis divided into separate rows for sum of squares reduced
(SSR, SSE(C)-SSE(A)), sum of squared errors for the compact model (SSE(C)) and sum of
squared errors for the augmented model (SSE(A))
Q6: when deciding whether to reject Model C, we want to know whether the calculate d value of
the PRE is a _____, assuming that Model C is correct; if so, do not reject
Answer: likely value
Q7: a PRE value is surprising when it occurs___?
Answer: less than 5% of the time when the null hypothesis is true
Q8: two reasons for calculating F
Answer: to examine the proportional reduction in error per parameter added to the model 2) to
compare the proportion of error that was reduced (PRE) with the proportion of error that
remains (1-PRE)
Q9: refers to the chance of making a type I error
Answer: significance level
, Q10: refers to the proportion of times that the correct decision is made when Model C is incorrect
(we reject the null); 1-β
Answer: power
Q11: practice understanding this; e.g., if η^2=0 (null hypothesis is true), at what level can we
expect 95% or more of the observed PREs to be below?
Answer: PRE table
Q12: smallest effect size of interest
Answer: specifying an alternative hypothesis which would present a meaningful real-world deviation
from the null hypothesis
Q13: reduce error, increase α, increase n
Answer: ways to improve power
Q14: β0 = B0
Answer: null hypothesis; statement in which the true parameter of a population is equal to some
outside specific value, determined a priori
Q15: σ2
Answer: population variance; determines the spread of the normal curve, larger = wider
Q16: η^2 = 0
Answer: no true proportional reduction in error, Model C (the null hypothesis) is correct and not
rejected, Model A makes no further improvements
Q17: σ^2/n
Answer: equation for the variance of a sampling distribution of means (when drawn from a normal
distribution); narrower than original sampling distribution of observations, gets even
narrower as we collect more observations (because n increases, thereby increasing the
denominator)
Q18: 4
Answer: for most reasonable numbers of observations, the 95% critical value for F (required to reject
the null) is approximately 4
Q1: why is the mean the best estimator (in most cases) for a normal distribution?
Answer: more efficient and consistent
Q2: when using the mean and SSE, always--?
Answer: check for outliers
Q3: 1 --> n - p - how do you convert the simple model for the MSE into the complex model?
Answer: n
Q4: what is the question we consider when we're deciding whether we can reject the null
hypothesis (β0 = B0)?
Answer: is Model A better enough than Model C
Q5: analysis of variance (ANOVA) table
Answer: summary of the results of analysis divided into separate rows for sum of squares reduced
(SSR, SSE(C)-SSE(A)), sum of squared errors for the compact model (SSE(C)) and sum of
squared errors for the augmented model (SSE(A))
Q6: when deciding whether to reject Model C, we want to know whether the calculate d value of
the PRE is a _____, assuming that Model C is correct; if so, do not reject
Answer: likely value
Q7: a PRE value is surprising when it occurs___?
Answer: less than 5% of the time when the null hypothesis is true
Q8: two reasons for calculating F
Answer: to examine the proportional reduction in error per parameter added to the model 2) to
compare the proportion of error that was reduced (PRE) with the proportion of error that
remains (1-PRE)
Q9: refers to the chance of making a type I error
Answer: significance level
, Q10: refers to the proportion of times that the correct decision is made when Model C is incorrect
(we reject the null); 1-β
Answer: power
Q11: practice understanding this; e.g., if η^2=0 (null hypothesis is true), at what level can we
expect 95% or more of the observed PREs to be below?
Answer: PRE table
Q12: smallest effect size of interest
Answer: specifying an alternative hypothesis which would present a meaningful real-world deviation
from the null hypothesis
Q13: reduce error, increase α, increase n
Answer: ways to improve power
Q14: β0 = B0
Answer: null hypothesis; statement in which the true parameter of a population is equal to some
outside specific value, determined a priori
Q15: σ2
Answer: population variance; determines the spread of the normal curve, larger = wider
Q16: η^2 = 0
Answer: no true proportional reduction in error, Model C (the null hypothesis) is correct and not
rejected, Model A makes no further improvements
Q17: σ^2/n
Answer: equation for the variance of a sampling distribution of means (when drawn from a normal
distribution); narrower than original sampling distribution of observations, gets even
narrower as we collect more observations (because n increases, thereby increasing the
denominator)
Q18: 4
Answer: for most reasonable numbers of observations, the 95% critical value for F (required to reject
the null) is approximately 4
