A quantile-quantile (QQ) plot checks whether observed values are compatible with a normal distribution. It places standardized observed values, often model residuals, against theoretical normal quantiles. Points that closely follow the diagonal line support normality; notable departures, especially in either tail, suggest skewness or another mismatch with the normal pattern. section 5.7.5 section 17.4.2
QQ plots are useful both before modeling and after fitting a model. In exploratory data analysis, they provide a fast visual view of each continuous variable’s distribution. In regression and one-way ANOVA diagnostics, they help assess whether residuals meet the normality assumption needed to judge model integrity. section 14.5.2 section 17.4.3
Read a normal QQ plot by comparing its points with the diagonal reference line:
- Points mostly on the line indicate that the data or residuals are approximately normally distributed. section 5.7.5 section 17.4.3
- Small deviations near both tails can occur without raising serious concerns when the overall alignment remains close. section 5.7.5 section 14.5.2
- Large tail departures indicate that the values are not well aligned with a normal distribution and may be skewed. section 17.4.2
- Treat the plot as a visual clue rather than as a direct display of the residual distribution. section 14.5.2
Checking NBA model residuals
After fitting an ANOVA of standardized regular-season wins by scoring category, request the Residuals vs. Fitted and Normal QQ diagnostics. The QQ points deviate slightly at both tails but otherwise overlay the diagonal, which supports approximate residual normality. The accompanying residual-versus-fitted plot addresses a separate assumption: equal variance across groups. section 14.5.2
Screening point-spread variables
For NBA betting data, QQ plots can screen continuous variables during EDA. SmartEDA’s output identifies openspreadH as normally distributed while diff_margin_closespreadH is not; density plots provide a second view of the same contrast. section 17.4.3
plot_qq(oddsdf3)
Use DataExplorer’s plot_qq() call to generate a matrix of normal QQ plots for the continuous variables in a data frame. section 17.4.2
