LFD Book Forum (http://book.caltech.edu/bookforum/index.php)
-   Homework 2 (http://book.caltech.edu/bookforum/forumdisplay.php?f=131)

 geekoftheweek 01-18-2013 05:04 PM

numerical issues

On questions 5, 6, 7 I keep getting answers that are about half way between two possible values. The answers I'm getting make sense to me and because they are not *exactly* halfway between the two answers I am technically able to find a closest value, but that's just chance. I expect that the options are all way off except one right? But that's not what I'm seeing.

For instance, if there are two options: 1 and 10, I might get a value of 5. Yes, 5 is closer to 1 than 10 but not by much, and it's making me have no confidence in my answers.

If we are getting answers that are sort of on the fence like this does that imply that they are dead wrong? Please advise.

Thanks

 sanbt 01-18-2013 05:51 PM

Re: numerical issues

Quote:
 Originally Posted by geekoftheweek (Post 8840) On questions 5, 6, 7 I keep getting answers that are about half way between two possible values. The answers I'm getting make sense to me and because they are not *exactly* halfway between the two answers I am technically able to find a closest value, but that's just chance. I expect that the options are all way off except one right? But that's not what I'm seeing. For instance, if there are two options: 1 and 10, I might get a value of 5. Yes, 5 is closer to 1 than 10 but not by much, and it's making me have no confidence in my answers. If we are getting answers that are sort of on the fence like this does that imply that they are dead wrong? Please advise. Thanks
For 5 and 6, it's possible that the result is almost halfway between 2 answers, but it should always favor one answer.

 gah44 01-20-2013 02:39 AM

Re: numerical issues

Quote:
 Originally Posted by geekoftheweek (Post 8840) On questions 5, 6, 7 I keep getting answers that are about half way between two possible values. (snip) For instance, if there are two options: 1 and 10, I might get a value of 5. Yes, 5 is closer to 1 than 10 but not by much, and it's making me have no confidence in my answers. If we are getting answers that are sort of on the fence like this does that imply that they are dead wrong? Please advise. Thanks
If you repeat the experiment 1000 times, and average, it is more and more unlikely that you get the wrong closest answer.

If you are still unsure, do another 1000 times and verify that it is still the same.

For problems where the answer is an integer , the standard deviation is and the standard devation of the mean of trials is the standard deviation divided by , or .

To be reasonably sure, you should be a few standard deviations away from half way.

 All times are GMT -7. The time now is 12:24 AM.