Page: 1 2 3 4 5
Sezei
Registered user
Global user
Registered: 012017
Posts: 89
Karma: 1 (+3/2)

Reply  Quote


Re: Real IQ calculation
quote: Benne161 wrote:
@Sezei
Yes
Do you believe that scores above 130 or 145 in scientifically recognized tests be more accurate or IQ's in that range can be measured exactly at all?
quote: Sezei wrote:
Mmh complex question, don't have time the afternoon to answer. I'll try this evening !
PS : I'm french, so it's 3 P.M. here, just to be sure we're phased ^^
In the end I don't have time to invest in a huge answer now. As I explained in the other topic, I have emergencies.

7/17/2017, 9:52 pm

Link to this post
Email Sezei
PM Sezei
Blog

Benne161
Registered user
Global user
Registered: 062016
Posts: 238
Karma: 2 (+5/3)

Reply  Quote


Re: Real IQ calculation
Thats ok , no problem :)

7/17/2017, 9:54 pm

Link to this post
Email Benne161
PM Benne161
Blog

Blanka86
Registered user
Global user
Registered: 032017
Posts: 144
Karma: 7 (+7/0)

Reply  Quote


Re: Real IQ calculation
My real IQ calculated on IQEXAMS is now 145. It's more than I could wish for when I started in this world of highranking IQ tests.
Greetings.

7/31/2017, 11:31 pm

Link to this post
Email Blanka86
PM Blanka86
Blog

IQnavi
Head Administrator
Global user
Registered: 062016
Posts: 239
Karma: 13 (+13/0)

Reply  Quote


Re: Real IQ calculation
Hi,
me and Jakub have been working on a method to use the:
Ferguson's Formula  The Statistical Technique
for Combining IQ Scores
It is looking good and I think we can present both the rarity mean and this as two alternative "Real IQs", then you can judge what can is correct. (Hopefully they are very similar)
BT / Hans

8/4/2017, 6:56 pm

Link to this post
Email IQnavi
PM IQnavi
Blog

Rosberg06
Registered user
Global user
Registered: 062017
Posts: 48
Karma: 1 (+1/0)

Reply  Quote


Re: Real IQ calculation
quote: Rosberg06 wrote:
I think whether using Iq scores or rarities in the calculation, the average is not the appropriate tool to use when dealing with this kind of data. I believe the median is the suitable one.
Further, with enough data, you can design your own PDF and you can set a confidence interval for a score to fall within and make predictions about your score being equal or greater than a certain value, again using the median as tool instead of the average. Just my opinion.
I know it's a sensitive subject(perhaps more sensitive than I imagined) so I will try to rephrase. I still think that from a statistical point of view this is the way to do it. But there's no reason to assume that applying another algorithm is incorrect as long is based on strong arguments,and not only this, but it can also be better suited.
With this in mind I was curious enough to see if I can capture something more in my iq scores data. So I took a few hours to put together some simple statistics in an Excel spreadsheet.
Assuming my data is normally distributed, the mean, median were calculated, data intervals for one, two and three sigma deviation were set up as well as the confidence intervals at 95% and the probability of a given set to fall in an interval.
Also I tried a simple regression analysis between my first and second attempts at the tests, more precisely the 2nd attempts were regressed on the 1st attempts scores(the intention was to see how my 2nd attempts are influenced by 1st attempts movement).
Interesting enough is that for my 2nd attempt the 95% confidence interval of (143,148) is quite close to my 2nd attempt(with the upper limit matching) on JCTI which gave me a score of 145 with an (138, 148) 95% confidence interval.
Using the regression model the predicted 2nd attempt value for a 134 1st attempt(which is also my 1st attempt on JCTI) is 143(close enough to my 145 score in 2nd attempt on JCTI, altough I have a couple of tests with only one attempt so if I manage to raise my scores the prediction should get even closer).
Also, when using rarities in the calculation for the 2nd attempts (150 if I remember correctly), still a little bit outside the 99% confidence interval (141,149).
Edit: It's true that the regression model needs further analysis as more assumptions have to be made(e.g. error terms are independent and identically distributed and also normally distributed and succesive error terms are independent of each other, that is there's no autocorrelation) so more parameters should be checked. I calculated some, and with the correlation coefficient of [sign in to see URL] and Rsqured of [sign in to see URL] (i.e. the linear model accounts for 38% variability of the data) things aren't looking too bad. However more parameters should be investigated so better conclusions can be drawn (Excel has a static regression model where a lot more parameters are estimated but because I run Linux, hence an open version of Excel, I can do that).
Overall I think it's an interesting way to look at your collection of scores and pretty fun to play with. I wish I had more tests done as I only have 24.
And a picture..
Log in or sign up to see linked image content
Last edited by Rosberg06, 8/8/2017, 1:57 pm

8/8/2017, 1:49 pm

Link to this post
Email Rosberg06
PM Rosberg06
Blog

jakubnowak
Registered user
Global user
Registered: 082016
Posts: 100
Karma: 12 (+13/1)

Reply  Quote


Re: Real IQ calculation
Hi
thanx for sharing interesting
Jakub

8/10/2017, 6:38 am

Link to this post
Email jakubnowak
PM jakubnowak
Blog

Rosberg06
Registered user
Global user
Registered: 062017
Posts: 48
Karma: 1 (+1/0)

Reply  Quote


Re: Real IQ calculation
Hi Jakub,
My pleasure. I guess if anyone is curious enough to try with their own data, I'm willing to share the file.

8/10/2017, 9:56 am

Link to this post
Email Rosberg06
PM Rosberg06
Blog

Add a reply
Page: 1 2 3 4 5
You are not logged in ( login)
