So far we have just interpreted the estimated regression coefficients. However, remember that our dataset is just a sample of the population. So the regression coefficients that we get from the analyses on our sample are only estimates of the true population parameters. That's why they're called parameter estimates. For example, if we were to draw another sample from our population, and test the same regression model on the new sample. The parameter estimates are not likely to be the same as they were with our original sample. This is due to sampling variability, meaning that the sample we draw is not likely to be exactly like the population. To get a better understanding of how our sample estimates represent the population values, we can look at confidence intervals. Confidence intervals tell us which values of the parameter estimates are plausible in the population. Typically, we look at 95% confidence intervals, which tell us, with 95% certainty, the range of parameter estimate values, that includes the true population parameter. That is, we are 95% certain that the true population parameter falls somewhere between the lower and upper confidence limits that are estimated based on our sample perimeter estimates. Let's rerun the code for our multiple regression model, predicting nicotine-dependent symptoms. But first, we'll add some new code asking SaaS to also print 95% confidence levels. We do this by adding the clparm option after solution. Now if you take a look at the output, you'll see that the results are the same although there is an additional set of columns for the parameter estimates that provide the estimated lower and upper limit for the 95% confidence interval for each parameter estimate. For example, if we take a look at the parameter estimate for Major Dep Life. We see that it is 1.3. Meaning that on average, individuals with major depression have 1.3 more nicotine dependent symptoms than people without major depression. This is our point estimate of the population parameter. If we look at the conference interval though, we see that it ranges from 1.1 to 1.5, meaning that we're 95% certain that the true population parameter for the association between major life depression and number of nicotine dependent symptoms fall somewhere between 1.1 and 1.5. That is, in the population there's a 95% chance that people with major life depression have anywhere between 1.1 and 1.5 more nicotine dependent symptoms than people without major depression. Also note that our dysthymia variable this life, had a P-value of .19, which is not statistically significant. We could not reject the null hypothesis of no association between dysthymia and number od nicotine dependent symptoms afterly adjusted for major life depression and the other explanatory variables in the model. If we take a look at the confidence interval for this variable. We see that it ranges from -.1 to .7. Which includes a value of zero in that range. This means that we can't rule out with 95% confidence the possibility that the association between dysthymia and number of nicotine dependent symptoms, after adjusting for other variables in the model, is zero. In linear regression, when you have a nonsignificant P value, the 95% confidence interval for the parameter estimate will include a value of 0, no association.