PakMediNet Discussion Forum : Biostatistics : 95% Conf. Int
95% Conf. Int
An researcher reported that among 100 patients 63 patients (63%, 95% CI 52.8-72.4%) were suffering from the disease.
What is the meaning of 95%CI, its importance, its proper name in statistics; and how it is calculated using SPSS.
If anonter researcher reported that among 100 patients 63 patients (63%) were suffering from the disease. WHAT WOULD BE THE DIFFERENCE.
THANKS
Posted by: tariqaslPosts: 42 :: 07-04-2010 :: | Reply to this Message
Re: 95% Conf. Int You normally calculate any characteristics for population which is based on sample observations, so there is always doubt about exact value of population, when you calculate 95% CI for parameter , then you are confident that results will favour for 95% or you are 95% sure that population parameter lies between two values. for more any free help call 0300-4668681
Posted by: ibrahim_apPosts: 138 :: 09-04-2010 :: | Reply to this Message
Re: Re: 95% Conf. Int Thanks; but how to calculate it using SPSS??
Posted by: tariqaslPosts: 42 :: 11-04-2010 :: | Reply to this Message
Re: 95% Conf. Int
I have been corrected by a respected friend and statistician on this very forum that the definition of 95% CI is different depending upon the statistical framework one uses.
In classical statistics or frequenstist statistics (this is the one that we see all around us) 95% CI means that if an experiment is repeated many many times in exactly the same manner, 95% of the values of the parameter will be in that 95% CI.
Strictly speaking, frequentist statistics doesn't talk about the 95% confidence that the mean is in that interval. This interpretation is based on Bayesian statistical framework (you may have not seen it used). In Bayesian statistical framework, 95% CI is also called 95% credible interval and its is interpreted as the interval in which we are 95% sure that the true population parameter value is present (this depends on the available information at that time; you can update your 95%CI if you get additional information).
In my limited understanding, this difference in interpretation is due to the beliefs about the population parameter in each framework. In frequentist framework, a true population mean is a fixed value and not a random variable while in Bayesian statistic true population parameter is a random variable. Therefore, in Bayesian statistics one can have an uncertainty about true population mean but not in frequentist statistic.
Regarding question of how to calculate it in SPSS, I have only one simple answer; learn biostatistics - there is no royal road to knowledge.
Posted by: rqayyumPosts: 199 :: 12-04-2010 :: | Reply to this Message
Re: 95% Conf. Int
In SPSS version 16.0 there is a way to calculate this. Say for example 63 out of 100 guests at dinner party ate chicken.
Now come at the variable view (which is present in the lower left corner of the screen) and make a new variable. Name it as "atechicken" and mark it as a "numerical variable". Under the "measure" tab write it as "scale"
Now come in the data view and now for every guest who ate chicken put 1 and who didn't put a zero (0). When your data entry is completed then open the ANALYZE>DESCRIPTIVE STATISTICS>EXPLORE. In the explore tab you d see "confidence interval" and you can set it at 95 or at any other value.
Hit OK and you d get the result.
I m not 100% sure that its the only way but by banging my head several times with a wall I came up with this.
@ rqayyum. Sir you r very right regarding this but spare a thought for us poor fellows who dont study a dime of biostatistics in our undergraduate years and those of us who are fortunate enough to have attended the computer CPSP sponsored workshop know that there is a 15 minute class (yes I m not kidding) on how to use the SPSS software. I guess none of us understands even how to handle the basic data entry.
Then at most of the workplaces biostatician presence is a non-entity. It is impossible for most of us to even complete our research proposals during our clinical postings.
What remedy would you suggest to an average doctor?
Regards.
Posted by: doctocPosts: 20 :: 07-05-2010 :: | Reply to this Message
Re: Re: 95% Conf. Int
VERY NICE COMMENTS AND INPUT. KEEP IT UP. THANKS
Posted by: tariqaslPosts: 42 :: 21-05-2010 :: | Reply to this Message
Re: 95% Conf. Int
Dear @doctoc, I understand your frustration. The way you calculate 95%CI confidence interval depends on many things, most importantly the parameter you are trying to estimate and the method that you are using to estimate your parameter. Unless one has understanding of what one is doing, it is very likely that one will have incorrect 95%CI. Yes, you will have some numbers but whether those are correct or not, that is a different matter.
What you are getting in your example is the 95%CI for a proportion (your input is dichotomous). This will not be the way to get 95%CI for other types of estimates.
Posted by: rqayyumPosts: 199 :: 07-06-2010 :: | Reply to this Message
Re: 95% Conf. Int Sorry to say; still i am unable to calculate it.
Posted by: tariqaslPosts: 42 :: 10-06-2010 :: | Reply to this Message
Re: Re: 95% Conf. Int You will have to give much more detail than what you have given so that someone can help you. Describe your study question, provide point estimates and dispersion of your data, and what you are trying to do.
Posted by: rqayyumPosts: 199 :: 22-06-2010 :: | Reply to this Message
Re: Re: 95% Conf. Int Dear Doctor, dont worry, please contact me at my email address, i will solve your problem, and what u actually need or want. ibrahim_ap98@yahoo.com
Posted by: ibrahim_apPosts: 138 :: 25-06-2010 :: | Reply to this Message
Re: Re: Re: 95% Conf. Int
I have 200 sample size; the prevalence of X problem is in 50 patients i.e. 25%. Now the reviwer says that dont mentions the 25% only, but also mention the 95%CI values as well.
Now how to calculate it?
I have given example of X, But i have to calculate prevalence of many other problems.......
Plz help---
Thanks
Posted by: tariqaslPosts: 42 :: 01-07-2010 :: | Reply to this Message
Re: Re: Re: Re: 95% Conf. Int
for your results below is 95% CI for 25%
Exact 90% C.I. (Mid-P) = 0.202 to 0.303
Exact 90% C.I. (Fisher's) = 0.200 to 0.305
Exact 95% C.I. (Mid-P) = 0.194 to 0.314
Exact 95% C.I. (Fisher's) = 0.192 to 0.316
Exact 99% C.I. (Mid-P) = 0.177 to 0.334
Exact 99% C.I. (Fisher's) = 0.175 to 0.337
For any details feel free to call 0300-4668681
Posted by: ibrahim_apPosts: 138 :: 03-07-2010 :: | Reply to this Message
Re: Re: Re: Re: Re: 95% Conf. Int
I am very thankful to you, but what is the method to calculate
and how to calculate these in SPSS?
Thanks for help
Posted by: tariqaslPosts: 42 :: 06-07-2010 :: | Reply to this Message
Re: Re: Re: Re: Re: Re: 95% Conf. Int
I don't use SPSS so cant help there, perhaps someone else may know. While there are several methods for calculating 95%CI of a proportion (Exact, Wald, Wilson, Agresti, Jefferys) they all give about the same values.
The easiest method (and you can do it by hand) is as below:
1. calculate standard error by using
sqrt(p(1-p)/n)
2. Multiply this standard error with 1.96 (z-value for 95%CI)
3. Add (or subtract) the resulting number to your proportion; this will give you upper (or lower) limit of 95%CI
Here p is proportion which in your case is 0.25; n is your sample size.
Please note, as I have stated earlier, methods for calculating 95%CI vary depending upon your data and research question and the above method will not work in most other cases; however, it will work for calculating 95%CI of any one proportion.
[Edited by rqayyum on 13-07-2010 at 05:47 PM GMT]
Posted by: rqayyumPosts: 199 :: 13-07-2010 :: | Reply to this Message
Re: 95% Conf. Int in spss you just go in descriptive statistics and then go right side button statistics there u will definitely find the confidence int ok
Posted by: RaliPosts: 5 :: 06-08-2010 :: | Reply to this Message