Register to reply 
Probability of 0 bit in ASCII text files 
Share this thread: 
#19
Oct912, 06:38 AM

P: 54

Still confused, say, why getting a 0 eliminates 7 of the total weight of 16 etc?.(saying it in simple: we have Pr(MSB=0)=1/8, plus Pr(0)=Pr(1)=1/2, then draw a bit, (as whole) Pr(0)=?, Pr(MSB=0)=?)
Anyway, it seems to follow the rule, where i equals number of 0 drawn sequenced/continual, say i=0,1,2. 1/8 ((i+1)/2^i +(7i)/2^(i+1))=(i+9)/2^(i+4) . Agree? 


#20
Oct912, 06:41 AM

P: 54




#21
Oct912, 03:50 PM

Homework
Sci Advisor
HW Helper
Thanks
P: 9,877

Let's try it using the standard rules of conditional probability.
P[bit was MSB  bit was 0] * P[ bit was 0] = P[bit was MSB & bit was 0] = P[bit was MSB] = 1/8. P[ bit was 0] = 9/16 (you already proved) So P[bit was MSB  bit was 0] = (16/9)*(1/8) = 2/9 


#22
Oct1012, 03:19 AM

P: 54

So you agree that the Prob of drawing a 0 bit in ASCII is Pr(0)=9/16 (given each bit is randomly distributed except MSB)? Then, I still wonder about Pr[00]=? (I try to say that Prob of drawing a 2nd 0, given 1st was 0 too). Because I was told 9/16*9/16 comes close but not correct. 


#23
Oct1012, 12:41 PM

Homework
Sci Advisor
HW Helper
Thanks
P: 9,877




#24
Oct1012, 11:02 PM

P: 54

I was told it is 10/32 (drawing two zeroes with prior info), and it follows the rule of 1/8 ((i+1)/2^i +(7i)/2^(i+1) ). Just can`t figure out.



#25
Oct1112, 02:45 AM

Homework
Sci Advisor
HW Helper
Thanks
P: 9,877

You already know prob that first was a 0 was 9/16. Using the joint probability rule, you can easily calculate the prob that second is a zero given the first was. Do you see how? 


#26
Oct1112, 04:18 AM

P: 54

PLEASE!!!!!!!!!!! 


#27
Oct1112, 11:50 AM

Homework
Sci Advisor
HW Helper
Thanks
P: 9,877

OK. First we need the probability that N successive bits are all zero. For now, assume N < 9. What is the probability that the N include an MSB? N/8, right?
If they do include an MSB, what is the prob they are all 0? There must be exactly one MSB, so it's 2^{(N1)} ok? And if they don't, it's 2^{N}. So in total 2^{N}(1+N/8). Now suppose we know the first N1 were all zero and we want the prob that the Nth is too. By the conditional/joint probability rule, we can just take the ratio of two of these probs: 2^{N}(1+N/8)/(2^{(N1)}(1+(N1)/8)) = (8+N)/(2(7+N)) Do you also need to investigate N > 8? 


#28
Oct1812, 04:34 AM

P: 54

Thanks, let me understand your explanation first. I think it takes sometime!



#29
Nov2212, 05:12 AM

P: 54

3/8*1/2 + 5/8*1/2*1/2 = 11/32. Since you computed of having 2 successive 0 bits is 10/32. So my calculation of (000) seems wrong? 2nd : if it follows to the rule of (2^N)(1+N/8), Then 5 successive 0 bits(00000) would be: 2^5(1+5/8)= 13/512 So my calculation of (000) seems wrong. Will you please share your thoughts? 


#30
Nov2212, 03:04 PM

Homework
Sci Advisor
HW Helper
Thanks
P: 9,877




#31
Nov2512, 03:34 AM

P: 54

Please let me confirm about the "1/2". 1st. Analysis: The number of "1/2 " at the right part (of +) is 2^(N1), and left is 2^N. right? but Where the " number of 1/2 " comes from? Is that because of the N that is number of successive zero bits taken from ASCII? 2nd. If N>8, then we should assume it is impossible. In other words, the answer should be 0. Am I correct? 


#32
Nov2512, 02:05 PM

Homework
Sci Advisor
HW Helper
Thanks
P: 9,877

With probability N/8, there is a leading bit. That bit will be 0. The other N1 may be 0 or 1, equally likely. So prob that all N are 0 is 2^{(N1)}. With prob 1N/8, there is no leading bit. All N may be 0 or 1, equally likely. So prob that all N are 0 is 2^{N}. Adding up: (N/8)*2^{(N1)} + (1N/8)*2^{N} = 2^{N}(1+N/8). 


#33
Nov2512, 06:44 PM

P: 54

So in other words, 「We can assume that there are seven bits (or N<8) and that there is at most one MSB bit (which means either one MSB or zero MSB) in it. Thus we can compute the probability of having zero MSB, plus the probability of having one MSB. right ?」 1st. as the same token, If we let N be 15bits , then the it must contain at least one MSB or at most two MSBs. We add the probability of having one MSB and having two MSBs in it. 2nd. further, when N is 37 bits, then it must contain at least four or at most five MSBs. And so on ... So what it turns out to be? Just replacing N with 8A+B seems not sufficient for the computations, unless I understand what 'A', 'B' stand for. Will you furnish some explanations further please? 


#34
Nov2512, 07:30 PM

Homework
Sci Advisor
HW Helper
Thanks
P: 9,877




#35
Nov2512, 08:37 PM

P: 54

P(MSB)=P(MSB H0)P(H0) + P(MSB H1)P(H1) (Let Hi be the event that there are i MSB bits in N, for i = 0, 1, 2, 3….. ) where P(MSB H0) stands for conditional probability of MSB bit in N given it is H0 which equals 0 (no MSB) and P(MSB H1) stands for conditional probability of MSB bit in the N given it is H1 which equals 1/7 (one MSB); But it is little different from the point of N=8A+B. right? I could not figure out the computation. 


#36
Nov2512, 11:08 PM

Homework
Sci Advisor
HW Helper
Thanks
P: 9,877

Are we done with the N≤8 case? 


Register to reply 
Related Discussions  
How read text files on CASIO fx9860GII SD?  Calculators  2  
Several ascii files, read, write, average?  Programming & Computer Science  5  
Read multiple binary files to ascii  Programming & Computer Science  0  
Fortran 90 question about reading files with text  Programming & Computer Science  1  
Indexing a lot of text/pdf files  Computing & Technology  1 