стр. 63 |

explained by the

Below 1. A higher

square Fit. Superior to R-square

independent

value is better because it is sensitive to

variables or R-

the addition of irrelevant

square adjusted

variables.

for # of

independent

variables

Look at the p- For a one-tailed test (at

value (in the 95% confidence level), the

-Ratios The reliability of column вЂњSig.вЂќ) it critical value is

our estimate of the must be low: (approximately) 1.65 for

individual beta testing if the coefficient is

- below .01 for greater than zero and

99% confidence in (approximately) -1.65 for

222

Chapter 12: Regression

Name Of What Does It

Statistic/ Measure Or Critical Values Comment

Chart Indicate?

the value of the testing if it is below zero.

estimated

coefficient

- below .05 for

95% confidence in

the value of the

estimated

coefficient

- below .1 for 90%

confidence in the

value of the

estimated

coefficient

Confidence The upper and Any value within the

The 95%

Interval for lower values give confidence interval

confidence band

beta the 95% cannot be rejected (as the

for each beta

confidence limits true value) at 95% degree

estimate

for the coefficient of confidence

Charts:

Scatter of

Extremely useful for

predicted

checking for breakdowns

dependent There should be

of the classical

variable and no discernible

assumptions, i.e. - for

residual pattern. If there

problems like mis-

(Preferably is a discernible

specification and/or

after Mis-specification pattern, then do

heteroskedasticity. At

standardizing and/or the RESET and/or

the top of this table, we

heteroskedasticity

the series) DW test for mis-

mentioned that the F-

specification or

statistic is the first

Make a the WhiteвЂ™s test

output to interpret.

scatter chart for

Some may argue that the

manually heteroskedasticity

PRED-RESID plot is

after running

more important.

the regression

in Excel. **

Charts: plots There should be Common in cross-

of residuals no discernible sectional data.

against pattern. If there

independent Heteroskedasticity is a discernible If a partial plot has a

variables. pattern, then pattern, then that

(Preferably perform a formal variable is a likely

after test. candidate for the cause of

standardizing

223

Statistical Analysis with Excel

Name Of What Does It

Statistic/ Measure Or Critical Values Comment

Chart Indicate?

the series) heteroskedasticity.

Make a

scatter chart

manually

after running

regression**

Charts:

Histograms of

A good way to observe the

residuals. No

actual behavior of our

need to

Provides an idea The distribution residuals and to observe

standardize.

about the should look like a any severe problem in the

residuals (which would

distribution of the normal

Make an area

residuals distribution indicate a breakdown of

chart after

the classical

running the

assumptions)

regression in

Excel**

** (a) Estimate the series вЂњpredictedвЂќ by using the regression formula:

Predicted_Y= constant + B1X1 +вЂ¦ + BkXk.

(b) Standardize the series of predicted values using the function

INSERT/ FUNCTION/ STATISTICAL/ STANDARDIZE.

стр. 63 |