Statistical Sig. of Intervals

[brittanyharb]

Okay, so for determining whether the confidence interval is statistically significant, would the interpretations of the findings be similar to that of the R^2 test, where a number closer to one would be considered significant?

Like in #2, my interval for class size has a range of ~0.7. For class size, this shows a number that is less that a 1 student change in class size to be within a range of the 95% confidence interval. Would the interpretation of this be that this is not statistically significant because in order to meet the 95% confidence interval, the class size can only vary by 0.7, or is it the other way around?

[lecy]

The null is always that b1=0, which is a flat line representing no relationship between X and Y:

[lecy]

Saying that a regression model is statistically significant is equivalent to saying the confidence interval does not contain the NULL, i.e. that it does not contain zero.

[brittanyharb]

Ah I See! So this WOULD be significant because the interval of .7 is greater than 0 and the outcomes for the upper and lower calculations are on the same side of 0?

[lecy]

The interval can be on either side of the NULL. It just can't contain the the null, otherwise we can't say for certain if the program will have a positive or negative effect on the outcome.

[lecy]

At a bare minimum our evaluations must generate UNBIASED (accurate) estimates of program impact. Ideally our evaluations will also generate precise estimates. We will discuss this idea more over the next couple of weeks as we dive into control variables.

[brittanyharb]

Great! Thank you for the clarification