Scribd Logo
Upload

Welcome to Scribd!

  • 142 views

    Week 15: and and

    Uploaded by

    api-465329396
    (1) The formula for the variance of predictions for data outside the sample (Var(Y0 - Ŷ0)) is larger than the variance of predictions for data inside the sample (Var(Yi - Ŷi)). (2) The variance of predictions increases for data outside the sample compared to inside because data outside does not influence the regression coefficients in the same way. The covariance between inside data and coefficients leads to lower variance for inside predictions. (3) In summary, the variance of errors is expected to be smaller for predictions made on data used to construct the regression model compared to new data not part of the original sample.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    Week 15: and and

    Uploaded by

    api-465329396
    142 views1 page
    (1) The formula for the variance of predictions for data outside the sample (Var(Y0 - Ŷ0)) is larger than the variance of predictions for data inside the sample (Var(Yi - Ŷi)). (2) The variance of predictions increases for data outside the sample compared to inside because data outside does not influence the regression coefficients in the same way. The covariance between inside data and coefficients leads to lower variance for inside predictions. (3) In summary, the variance of errors is expected to be smaller for predictions made on data used to construct the regression model compared to new data not part of the original sample.

    Original Title

    79b4f705694535203382

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    (1) The formula for the variance of predictions for data outside the sample (Var(Y0 - Ŷ0)) is larger than the variance of predictions for data inside the sample (Var(Yi - Ŷi)). (2) The variance of predictions increases for data outside the sample compared to inside because data outside does not influence the regression coefficients in the same way. The covariance between inside data and coefficients leads to lower variance for inside predictions. (3) In summary, the variance of errors is expected to be smaller for predictions made on data used to construct the regression model compared to new data not part of the original sample.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    142 views1 page

    Week 15: and and

    Uploaded by

    api-465329396
    (1) The formula for the variance of predictions for data outside the sample (Var(Y0 - Ŷ0)) is larger than the variance of predictions for data inside the sample (Var(Yi - Ŷi)). (2) The variance of predictions increases for data outside the sample compared to inside because data outside does not influence the regression coefficients in the same way. The covariance between inside data and coefficients leads to lower variance for inside predictions. (3) In summary, the variance of errors is expected to be smaller for predictions made on data used to construct the regression model compared to new data not part of the original sample.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    You are on page 1of 1

    第 13 組

    Week 15
    Question 1

    For a multiple regression line, it can be shown in a matrix format 𝑌̂ = 𝑋𝑏 and 𝑌


    ̂𝑖 = 𝑋𝑖 𝑏 which is for a

    specific data 𝑋𝑖 inside the sample. As for data 𝑋0 outside the sample, the relation is 𝑌̂0 = 𝑋0 𝑏.

    ̂𝟎 ), 𝐕𝐚𝐫(𝒀𝟎 − 𝒀
    (1) Please use the above variables to find the formula for 𝐕𝐚𝐫(𝒀 ̂𝟎 ), and 𝐕𝐚𝐫(𝒀𝒊 − 𝒀̂𝒊 ),

    with Var(𝑌𝑖 ) = 𝜎 2 .

    (2) With the three formulas, please give some observations.

    Answer 1

    ̂𝟎 ) = Var(𝑋0 b) = 𝑋0 Var(b)𝑋0𝑇 = 𝑋0 ∙ (𝑋 𝑇 𝑋)−1 𝜎 2 ∙ 𝑋0𝑇 = 𝝈𝟐 (𝑿𝟎 (𝑿𝑻 𝑿)−𝟏 𝑿𝑻𝟎 )


    (1) 𝐕𝐚𝐫(𝒀

    ̂𝟎 ) = Var(𝑌0 − 𝑋0 b) = Var(𝑌0 − 𝑋0 ∙ (𝑋 𝑇 𝑋)−1 𝑋 𝑇 𝑌)


    𝐕𝐚𝐫(𝒀𝟎 − 𝒀

    = 𝜎 2 + 𝑋0 ∙ (𝑋 𝑇 𝑋)−1 𝑋 𝑇 𝜎 2 𝑋(𝑋 𝑇 𝑋)−1 ∙ 𝑋0𝑇 = 𝜎 2 + 𝜎 2 ∙ 𝑋0 ∙ (𝑋 𝑇 𝑋)−1 ∙ 𝑋0𝑇

    = 𝝈𝟐 (𝟏 + 𝑿𝟎 (𝑿𝑻 𝑿)−𝟏 𝑿𝑻𝟎 )

    𝐕𝐚𝐫(𝒀𝒊 − 𝒀̂𝒊 ) = Var(𝑌𝑖 − 𝑋𝑖 b) = Var(𝑌𝑖 − 𝑋𝑖 ∙ (𝑋 𝑇 𝑋)−1 𝑋 𝑇 𝑌)

    = 𝜎 2 + 𝑋𝑖 ∙ (𝑋 𝑇 𝑋)−1 𝑋 𝑇 𝜎 2 𝑋 (𝑋 𝑇 𝑋)−1 ∙ 𝑋𝑖𝑇 − 2 ∙ 𝜎 2 ∙ 𝑋𝑖 ∙ (𝑋 𝑇 𝑋)−1 𝑋𝑖𝑇

    = 𝜎 2 + 𝜎 2 ∙ 𝑋𝑖 ∙ (𝑋 𝑇 𝑋)−1 ∙ 𝑋𝑖𝑇 − 2 ∙ 𝜎 2 ∙ 𝑋𝑖 ∙ (𝑋 𝑇 𝑋)−1 𝑋𝑖𝑇

    = 𝝈𝟐 (𝟏 − 𝑿𝒊 (𝑿𝑻 𝑿)−𝟏 𝑿𝑻𝒊 )

    ̂𝟎 ) and 𝐕𝐚𝐫(𝒀𝒊 − 𝒀̂𝒊 ), we can find that the variance of data outside the sample
    (2) Comparing 𝐕𝐚𝐫(𝒀𝟎 − 𝒀

    is larger than that inside the sample. The reason is related to the covariance between data and

    coefficients of the regression line. Because the data 𝒀𝒊 inside influences the value of coefficient 𝐛, the

    variance of errors between 𝒀𝒊 and 𝒀̂𝒊 is expected to be less than that between 𝒀𝟎 and 𝒀
    ̂𝟎 .

    You might also like