Which statistical function should i use to approximate a forgeting curve?

Question

I have some "fuzzy" information how much someone still knows (or forgets) in reference to the time-dimension. Which (fuzzy?) function should i use to approximate the his forgeting curve? Would be a non-linear regression model enough?

ok, if i, for example, take ebbinghaus, how can i put my empiric data in this formula to approximate the parameters? Could someone post an example?

If i know, that i forget 50% after 1h, and 70% after 5h and so on.. how can i use this data to put it in the formula seen on wiki? (see post for the link of stroopguy)

Answer 1

I don't know, but I picture it as a line that at first descends gently, then geometrically accelerates in its angle of descent in relation to the "time" coordinate. This, in fact, seems to be a fit approximation of entropy in general, memory loss being, to my mind, a form of entropy.

This applies to any process that at first happens a little at a time, then accelerates in speed with the passage of time.

[But, if that won't work, you might try a sine curve, which appears to me the curve most commonly used by statisticians. They trot it out for almost every purpose.

There seems to be a correlation at work here: the more ignorant and gullible the audience, the more quickly a statistician will revert to the sine curve! The sine curve in the hands of statisticians is the statistical analogue of preachers yelling like Hell in the absence of a strong argument.

Sine curves are easy to draw, easy to picture, incomprehensible but impressive to most people, and, above all, safe! I have never heard of an argument being lost through use of a sine curve. Gentle curved lines are reassuring somehow.]

All of this is just my obscure way of guessing about something about which I don't have a clue. I get more humble all the time, but have not yet developed the ability to easily say "I don't know, Jack!" when I don't know.jack.

Answer 2

If Smith & Wesson makes weapons, and in turn, a criminal shoots somebody with a Smith & Wesson gun, then by ability of utilising reason and result, Smith & Wesson shot somebody. mutually as the two circumstances do have, what are stated as, "background reasons" in straight forward, the end result even nonetheless at as without delay as stems from an more suitable instant reason (devil and the criminal). that's likewise like a man or woman who beats his/her new baby. The abuser might have been abused as a clean baby, so, by ability of utilising background reason, we could in certainty blame the abuser's mom and pa. regardless of the incontrovertible fact that, the abuser ought to even nonetheless be held in fee. yet yet another difficulty: the region I stay, the community government is commencing as much as carry mom and pa in fee for teen eating. it fairly is to not assert that the infants are transforming into off without punishment, even nonetheless the background reason is likewise being addressed. in diverse words, you will be able to desire to pin blame on background reasons, even nonetheless the instant reason ought to get carry of an excellent type of the blame.

Answer 3

You need an exponential curve.

y=Ae^(kx)

Where:
y is the amount you still remember
A is the amount you originally remembered
k is a negative number
x is the independent (time) variable

Answer 4

Check out Wikipedia to see if this helps...

http://en.wikipedia.org/wiki/Forgetting_curve

Answer 5

Depends on what your modeling and what your "information" is. Will this be a stochastic process ?