confidence intervals
Locked
 
|
if I have a mean and a standard deviation of a variable, then I can say I am 95% confident the mean will not go up by more than 1.645 * StDEV.
can someone plz tell me what this 1.645 number would be for 98 and 99 % confidence. thanks so much. |
|
|
does it ever get as high as 4/5 standard deviations, that I could say maybe im 99.5% confident the mean will not go up by more than 4/5 standard deviations.
|
|
|
In reply to this post by jimjohn
You should be able to find these numbers in a standard z table.
Sara House [hidden email] >>> jimjohn <[hidden email]> 03/03/08 10:35 AM >>> if I have a mean and a standard deviation of a variable, then I can say I am 95% confident the mean will not go up by more than 1.645 * StDEV. can someone plz tell me what this 1.645 number would be for 98 and 99 % confidence. thanks so much. -- View this message in context: http://www.nabble.com/confidence-intervals-tp15807790p15807790.html Sent from the SPSSX Discussion mailing list archive at Nabble.com. ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD |
|
|
JimJohn,
Are you basing your estimate on a sample of the population? or on the population itself? What do you mean by "the mean will not go up"? Is there any chance that the mean could go down? These things matter. King Douglas American Airlines Customer Research Sara House <[hidden email]> wrote: You should be able to find these numbers in a standard z table. Sara House [hidden email] >>> jimjohn 03/03/08 10:35 AM >>> if I have a mean and a standard deviation of a variable, then I can say I am 95% confident the mean will not go up by more than 1.645 * StDEV. can someone plz tell me what this 1.645 number would be for 98 and 99 % confidence. thanks so much. -- View this message in context: http://www.nabble.com/confidence-intervals-tp15807790p15807790.html Sent from the SPSSX Discussion mailing list archive at Nabble.com. ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD |
|
|
In reply to this post by jimjohn
Not quite sure I understand the question!
The confidence interval is defined in relation to the population parameter; it is a (normally) symmetrical range of values with a stated probability of containing the population value (in this case, the population mean), estimated from a sample dawn from this population. Hence it involves the standard error rather than the standard deviation. Moreover, the population parameter is fixed, and it is not therefore clear what is meant by its 'going up'. If the population parameter changes it is no longer the one that you were estimating! The sample mean may change (e.g. in a before-and-after experiment), in which case you would want a confidence interval for the mean such change in the population, which again would be a question of estimating a fixed quantity. Regards, Julius > if I have a mean and a standard deviation of a variable, then I can say I > am > 95% confident the mean will not go up by more than 1.645 * StDEV. > > can someone plz tell me what this 1.645 number would be for 98 and 99 % > confidence. thanks so much. > -- > View this message in context: > http://www.nabble.com/confidence-intervals-tp15807790p15807790.html > Sent from the SPSSX Discussion mailing list archive at Nabble.com. > > ===================== > To manage your subscription to SPSSX-L, send a message to > [hidden email] (not to SPSSX-L), with no body text except the > command. To leave the list, send the command > SIGNOFF SPSSX-L > For a list of commands to manage subscriptions, send the command > INFO REFCARD > ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD |
|
|
In reply to this post by King Douglas
Ok well heres what I'm doing. I have a variable rate and another variable incentive. i do linear regression with rate as the dependent variable. now what im trying to do is estimate how much of the rate is completely not affected by changes to incentive. ie maybe the rate overall is 18%, but only 4/5% is not affected at all by incentive. so what i plan to do is see what will happen if incentive goes down by some number of standard deviations from the mean, then i plug that new incentive value into the regression equation to get a conservative estimate of my core rate not affected at all by incentive. the incentive could go up too, but i just want to estimate what will happen when it goes down. if it goes up, the rate will be even more. so i know from what was done before that if i lower the rate by 1.645 stdev, and then find my rates this will be 95% confidence. plz correct me if that is wrong or any toher advice. thx.
|
|
At 04:51 PM 3/3/2008, jimjohn wrote:
>I have a variable 'rate' and another variable 'incentive'. I do >linear regression with 'rate' as the dependent variable. I'm trying >to estimate how much of the rate is completely not affected by >changes to incentive. I.e., maybe the rate overall is 18%, but only >4/5% is not affected at all by incentive. Do you mean, the rate would be 4-5% in the absence of any incentive? >I plan to see what will happen if incentive goes down by some number >of standard deviations from the mean, then I plug that new incentive >value into the regression equation to get a conservative estimate of >my core rate not affected at all by incentive. If you mean what I thought you mean, above, the estimate of the 'core rate', i.e. what it would be without incentive, is the constant in the regression model. BUT, don't believe the estimate unless you actually data where 'incentive' is near zero. Otherwise, it's an extrapolation, and extrapolating from regression models is not accurate. It's especially not accurate if you extrapolate to 0 value of the independent variable; a lot of real effects behave differently, possibly non-linearly, near 0. Now, this has nothing to do with standard deviation, standard error, or confidence intervals. But, is it what you meant? -- No virus found in this outgoing message. Checked by AVG. Version: 7.5.518 / Virus Database: 269.21.7/1324 - Release Date: 3/10/2008 7:27 PM ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD |
«
Return to SPSSX Discussion
|
1 view|%1 views
| Free forum by Nabble | Edit this page |