|
|||||||||||||
MS Excel
Frequently Asked Questions |
|
|
XY Scatter Plot
Plotting X,Y data with anything but scatter plot may produce
unintended plots, whereas scatter always plots X,Y data directly
Simple Regression
After plotting X,Y data with a scatter plot, right click select one of the
rendered point, select the Add Trimline from the pop-up context menu.
Typically choose Linear Type. An option for displaying
R2 (correlation) is available.
Analysis Pack
If this is not installed by default, Menu: Tools > Add-Ins ... Sometimes, adding this causes a weird VBA error! Only work around known at the time for this is to uninstall Excel, then reinstall with all options checked (i.e. NOT install on first use).
More Regression - Single
Prepare two columns of data (first row labels optional). Menu: Tools > Data Analysis, Regression, OK. Cell range Y (typically this is main variable), Cell range X (typically this is independent variable), Labels (optional for first row). Enable Residuals, Residual Plots, and other options as necessary, OK.
SUMMARY OUTPUT Regression Statistics Multiple R 0.874854404 R Square 0.765370227 Adjusted R Square 0.748610958 Standard Error 2.79954673 Observations 16 ANOVA df SS MS F Significance F Regression 1 357.9249085 357.9249085 45.66847192 .00000920184 Residual 14 109.7244665 7.837461895 Total 15 467.649375 Coefficients Snd Error t Stat P-value Lower 95% Upper 95% Lower 95.0% Upper 95.0% Intercept 17.5381565 4.064297893 4.315174962 0.000712329 8.821096734 26.25521627 8.821096734 26.25521627 Mortality 0.324626544 0.04803699 6.757845213 9.20184E-06 0.221597355 0.427655732 0.221597355 0.427655732 RESIDUAL OUTPUT Predicted Standard Observ Mortality Residuals Residuals 1 53.18 14.12 1.94 2 58.37 -5.87 -0.80 3 72.98 -4.88 -0.67 4 77.46 7.14 0.98 5 77.94 -12.84 -1.76 6 80.77 -8.57 -1.18 7 82.42 -0.72 -0.10 8 84.54 4.66 0.64 9 87.37 -8.47 -1.16 10 89.72 -1.12 -0.15 11 90.90 4.10 0.56 12 92.55 -5.55 -0.76 13 94.20 1.70 0.23 14 95.85 8.65 1.19 15 96.09 4.31 0.59 16 99.16 3.34 0.46 Charts ...
F value large than 10 means the correlation is statistical significant. Associated low p-value further corroborates this.
So for each increase in Mortality, 95% certain Temperature increases between 0.22 and 0.43 . Data points with Stadnard Residual > +/-2 are consider potential outlander data
Regression validity Tests
- straight line (data not curved)
- Residual Data Values when plotted collect around a straight line
(using Normal P-Plots)
- Test for presence of Constant Variance
(residual vs predicted values)
- passes 'runs' test for elimination of periodic oscillation
More Regression - Multiple
Prepare several columns of data (first row labels optional).
Menu: Tools > Data Analysis, Regression, OK.
Cell range Y (typically this is main variable),
Cell range X (typically this is several independent variables - the columsn must be adjacent),
Labels (optional for first row).
Typically choose Confidence Level of 95%.
Enable Residuals, Residual Plots, and other options as necessary, it is best
to New Worksheet Ply with a name, OK.
More Regression - Multiple Predicting
Here's the deal. Only the main variable is not known ahead of time
for a limited range going forward. All independent variables are known
into the future. So this turns into an exercise of using the past
coefficients to compute the predicted values using already known future independent
values.
Basic Correlation
Have your equal length data columns ready. Menu: Tools > Data Analysis,
Correlation. Select Column data.
StatPack
If this is not installed by default, Menu: Tools > Add-Ins, Browse, Open StatPlusV25.xla . If successful a new Menu: StatPlus now present. ...
Normal Probability Plot
Fancy Correlation
Have your equal length data columns ready. Menu: StatPlus > Multivariate
Analysis > Correlation Matrix. May need to select Use Range References.
Select Column data. Output, typcailly new sheet - make sure name is NOT empty.
OK.
Results. High Correlation values are typically the most topcial values.
Pearson Probabilities (p-value or probability) less than .05, ideally zero,
are most topical. Correlation values with corresponding high p-values are
unlikely to be statistically relevent.
Text Boxes (Drawing Menu)
Test Boxes can be placed such that they float beside notable areas
of worksheet. Easiest way to create a Text Box is to use the Text Box
tool on the Drawing Toolbar. May need to adjust View settings if this toolbar
is not visible by default.