The chart shows the difference between the measured and the fitted value for each measured value with filled green squares, in Chi units (χ, difference divided by the standard deviation). If this value is positive, then the measured count value is greater than its matching fitted count.
Due to the statistical nature of the measured counts, we never expect a perfect matching between measured and fitted counts. Chi value is the ideal measure of the difference, because it takes into consideration the statistical variance of each measured count value, thus shows the significance of the difference, instead of its absolute value. On the chart, the Chi=1 limit is designated by a green dashed line, while a red dashed line shows the Chi=3 limit.
As a rule of thumb, if the absolute value of the difference is less than 1, it is considered not significant (no further improvement is required). On the other hand, if it is greater than 3, it is highly improbable that the difference is purely from the statistical scattering of counts, and closer inspection is needed in the vicinity of that channel.
The normalized sum of squared Chi values gives the Normalized Chi-Squared value (RXSQ, Reduced Chi-squared, χ2), which describes the quality of the fit for a whole region, and displayed in the region label at the first channel of the region. If RXSQ is significantly higher than 1.0, a manual refinement of the region fit is advisable.