Question

"You have developed a new anti-anxiety drug and are interested in its efficacy. You have two groups of rats (N=20 per group). You inject the one group (group 0) with saline, and inject the other group (group 1) with your anti-anxiety drug. Your outcome variable is the amount of exploratory behaviour (‘locomotion’) measured in a novel environment (a mild stress) over 1 hour. "

I ran an independent-samples t-test in SPSS (a statistics program) on this data and the same test but with bootstrapping.

Somehow the results were that the data is insignificant with a p-value of 0.139, alpha of 0.05, but the 95% confidence interval was 6.967 - 61.128.

If the data is insignificant, should 0 not be a value within the 95% confidence interval?

How is it possible that the null value (0) is not within the confidence interval, but still have a rather large insignificant result?

Answer 1

Here we have p value is 0.139

Alpha value is 95% level of significance that is 0.05

Here we can use p value approach we reject Ho if p value is less than alpha value. That is result is significant if p value is less than alpha value.

Our result p value is greater than alpha value here result is not significant at given level.

Here we have confidence interval : we can do results from confidence interval also

If interval include null value result is not significant.

Here interval is 6.967 - 61.128

That is interval not include zero or null value hence result is significant.

p value approach gives not significant and confidence interval gives the significant result.

## Note :

Some time our result is significant by p value approach but confidence interval include null value or vice versa.

Here p value approach is preferred rather than confidence interval.

And statistician and researchers believe in p value approach rather than confidence interval.

Hence here result is not significant.