Вы находитесь на странице: 1из 3

# CHECKING ASSUMPTIONS

1. Independence of Errors

##  Accounted for in design

 To reduce this error, researcher should assign an ID number to each
participant so that when it is entered, it can’t be entered twice
 If the same data is entered twice, it can bias the data set.

2. Linearity of Residuals

##  Ensuring the residuals follow a linear direction

o Regression line should be straight, indicating it is in the middle of
the data-scores
 Seen on an RVF plot whereby “scatterplot” is tilted on a horizontal
 Is there relatively equal data above and below the regression line?

If linearity was violated, it wold look like this (or a variation of) …

##  regress IV DV || regress barsold humidity

 rvfplot, yline(0)
3. Constant Variance (Homoscedasticity) of Residuals

##  Assumption wants to ensure there is consistency of variance of residuals

along regression line (Left and right)
 To see this, we need to run an RVF plot
o Places scatterplot on the horizontal
 Regression line = 0
 Violation of homoscedasticity would indicate “fanning” of data whereby
data is not consistently varied along regression line
 We want equality(ish) of variance

##  regress IV DV || regress barsold humidity

 rvfplot, yline(0)

## RVF Plot to Check 2 Assumptions

 Linearity
 Homoscedasticity
4. Normality of Residuals

 Checked through:
o Histogram
o Shapiro-Wilk
Histogram:
1. Need to create new residual variable
2. Plot histogram
3. Data should be surrounding 0 mark
4. Compare with SWILK

SWILK
1. Check SWILK statistic and compare significance with Histogram

##  regress IV DV || regress barsold humidity

 predict newvar_resid, resid || predict barsold_resid, resid
 histogram newvar_resid, freq bin (10)
 swilk newvar_resid