Time Series Analysis - ACCA Management Accounting (MA)
YouTube video
70 Comments
E
Ernest·
In addition to my earlier question about the variables, if I'm forecasting for 2003 and 2004 quarterly data using linear regression, what values should I input for the independent variable (e.g., time period) to generate predictions before making final adjustments?
E
Ernest·
Hi sir! In the video you said to use regression analysis to forecast trend. But could you please tell how? What will we take as variable x and what will we take as variable y?
What i wanted to find out myself.
Kindly assist sir
A
A-R·
Hi John, are we expected to to know the limitations of the model? If so, In the exam could I just say that it may be inaccurate because it uses historical data? are there any other limitations I should be aware of? Thanks John.
J
John MoffatTutor·
Firstly, you cannot be expected to write anything - all of the questions will be multiple choice etc..
If anything is asked them most likely is just the reason for the two different ways of calculating the seasonal variations (i.e. whether or not to calculate them as %'s).
(The fact that you are using historical data is a limitation of any forecast, but it not a limitation of the method because any method of forecasting will be based on historical data.)
A
A-R·
I understand, Thank you for clarifying John.
F
Ferdinard·
In minute 14:02 of the time series analysis, how does 86, the average of 84.75 and 87.25 coincide with 82? If anything, it is closer to 87 than it is to 82. I do not understand in what respect the two figures coincide. Please elaborate
J
John MoffatTutor·
The average of 86 sits alongside the actual result of 82, and so they are comparable.
B
Brandi·
Hello.
For the final question, you brought down 91.35 instead of 95.35 which changes the end answer. Where you got 398.26 I got 400.26 which led to a decrease by 0.065 and then a sum of 400. Is that correct?
V
Venoth·
yeah I got the same. John wrote down the wrong figure by accident.
So the total average would be 400.26 (like you said)
To make it 400%, we'd have to reduce the averages by a total of 0.065% (0.26/4)
So I reduced Quarter 1 and 2's averages by 0.06%, and Quarter 3 and 4's averages by 0.07%
Final figures: 99.95%, 102.96%, 95.61%, 101.48%
T
Thenuka·
Dear John,
If the Question asks for the Average Seasonal Variances for the Quarters,
Should the answers be +0.06,+3.13,-3.94,+1.44 OR -0.1125,+2.9575,-4.1125,+1.2675?
Thanks!
J
John MoffatTutor·
The first (+0.06 etc) as explained in the lecture, if using the additive model.
J
judy·
sir, In 3rd quarters of 2000 while calculating seasonal variance averages, you have written 91.35%, instead of 95.35%. therefore if we use 95.35% for the 3rd quarters of 2000,the average quarters will be 100.01%,103.02%,95.68%,101.55% and the total is 400.26. therefore we need to subtract 0.26 to get 400 as total. we have to subtract 0.06 for the 1st and 2nd quarters and 0.07 for 3rd and 4th quarters. so the total of each quarters are, 99.95, 102.96,95.61,101.48 respectively
M
Makhmudjon·
Hello Sir!
I understand how we get -0.17, (0.69/4= 0.17) but I do not understand -0.18 in 24:16 minute.
Could you explain it, please?
Thank you for considering my request.
J
John MoffatTutor·
It is just rounding. Because 0.69/4 = 0.1725, and I rounded it to 0.17, then they won't add up to exactly 0.69. To make it happen I adjusted one of them by 0.18, so now they will all add up to 0.69.
P
Peter·
Thk u so much mr john for your explanation.
I want to know why the total avaerge must add up to 400 and 0 using the multiplicative and additive model respectively?
J
John MoffatTutor·
With the multiplicative method, the variations are all bit more or a bit less than a 100. On average they are 100 and if there are 4 of them then the total must be 400.
With the additive model that are all a bit more than zero or a bit less than zero, so the average is zero and the total should be zero.
P
Peter·
Thank you so much for your clear response
J
John MoffatTutor·
You are welcome :-)
M
Mannan·
you said in the video that to check that the centered averages are reasonably linear, we can put them on a graph and see if they are more or less liner but I want to ask could we not use correlation coefficient to check the linearity?
J
John MoffatTutor·
Yes - by all means.
M
Mannan·
Hi sir! In the video you said to use regression analysis to forecast trend. But could you please tell how? What will we take as variable x and what will we take as variable y?
A
Abiriyi·
Thank you John for the lecture, when you said in the lecture that if we "forecast the trend then adjust by relevant seasonal variation" what would we be trying to achieve? Would we be trying to also forecast what the actual sales would be?
Also when referring to the multiplicative model as an alternative to the addictive model, you made mention of a hypothetical sales increasing and it being ten times higher at the end as compared to the beginning, by that do you mean ten times sales increase when we compare the difference between the sales figure at the beginning of the tenth year as against the sales figure at the end of the tenth year? Or ten times sales increase when we compare the difference between the sales figure at the beginning of the first year as against the sales figure at the end of the tenth year.
Sorry for the long question I have been on this topic for some days now and I can't help but ask for clarity
J
John MoffatTutor·
Yes, we would be trying to forecast what the actual sales would be in the particular period.
If the sales are increasing then it would be reasonable to expect that the seasonal variation would increase as well. The additive model has the same seasonal variation whatever the underlying level of sales is. The multiplicative model means that with higher underlying sales we will adjust for a higher seasonal variation.
A
Abiriyi·
Thank you John, well understood
J
John MoffatTutor·
Great :-)
F
faiza·
you are so good at explaining from the bottom to the top it immensely helps in clearing my doubts and so thank you very much
J
John MoffatTutor·
Thank you :-)
M
MohamedSupporter·
Great. Explained very well.
J
John MoffatTutor·
Thank you for your comment :-)
A
Asif·
The formula to find out the Average trend is:
(Last trend - first trend) / (no. of set of data - 1)
I want to understand why we minus by 1.
J
John MoffatTutor·
Don't rely on learning formulae for this exam.
Suppose there are four observations: 1, 2, 3 and 4
There are three differences 1 to 2; 2 to 3, and 3 to 4.
A
Asif·
Thanks sir for the logic. This is what I was looking for.
J
Jennifer·
what is the reason the average needs to be brought down to zero? I don't understand this area.
J
John MoffatTutor·
Some quarters are above the average and some are below the average. Since the average is in the 'middle' the total of the differences above the average must be the same as the total of the differences below the average - so they should add up to zero.
M
Mannan·
when you say "some quarters are above the average and some are below the average" which average are you referring to?
J
Jennifer·
Under Multiplicative model, I think there is an error under 2000 q3. shouldn't it be 95.35% instead of 91.35%
J
John MoffatTutor·
Yes it should - it is a silly mistake of mine. However the printed answer in the lecture notes has it correct :-)
K
Kyle·
Hi John,
If using the results of example 1 to forecast using regression analysis, which sets of figures would be considered X and which would be Y?
Many thanks,
J
John MoffatTutor·
Y would be the sales, and X would be the quarters (numbered consecutively).
J
John MoffatTutor·
You are welcome :-)
J
John MoffatTutor·
I assume you mean the total of the seasonal variations. Each of them are a little above or a little below 100%. The average of all of them should be 100% and therefore the total of 4 of them should be 400%
A
ahmedirfun·
I don't understand why the sum of the averages have to add up to 0
J
John MoffatTutor·
It is not the averages that add up to zero. It is the differences from the averages that should add up to zero - some are below the average and some are above the average, but the average is in the middle.
J
Jessica·
Hi
I don't understand why the centred average of 86 is said to be near quarter 3 than 2?
Please could you explain to me
J
John MoffatTutor·
The 4-quarter average of the first four quarters is in the middle - so is in between quarters 2 and 3.
Similarly the 4 quarters average of the next four quarters is in the middle - so is between quarter 3 and 4.
When we centre them we take the average of the middle of quarters 2 and 3 and the middle of quarters 3 and 4, and this then corresponds to quarter 3.
T
Tabasum·
Hi,
In the multiplicative model, the total averages arrives to 400.26%,so do we need to reduce the .26% to arrive at 400%??
J
John MoffatTutor·
Yes :-)
T
Tabasum·
OK, thank you :)
J
John MoffatTutor·
You are welcome :-)
M
mirakeryo26·
Hi, I don't understand why total +0.69 must be substracted until it become zero? Where do -0.17 and -0.18 come from?
J
John MoffatTutor·
The differences are all differences about an average - some are above the average and some are below the average but because they are all difference about an average the net figure should be zero. As I explain in the lecture, the reason they don't exactly add up to zero is because the first two and the last two observations have nothing to compare with.
At the moment they add up to a total of +0.69, so to make the total add up to zero we need to subtract 0.69. We subtract an equal amount from each of the 4 quarters. However because 0.69/4 = 0.1725 and we are only working to 2 decimal places, we subtract 0.18 from one of them and 0.17 from the others (and it does not matter at all which one you subtract 0.18 from).
A
Allen·
How do you calculate average change in the trend?
J
John MoffatTutor·
The only reason for wanting to know the average change in the trend would be for forecasting for the future. To forecast the trend we would take the total change for the periods given in the question and divide by the number of periods to get the average change per period.
A
aramontshonyana·
I was able to find them....
?
R
ritwick·
The calculated percent of the 3rd quater of year 2000 is 95.35 but while taken down to calculate the average percent it was entered as 91.35. Therefore the average percent of the 3rd quater should be 95.38 and there should be a difference of 2 ?
J
John MoffatTutor·
Thanks. I will check and if I have made a mistake I will correct it.
However the idea should still be clear (and I think the printed answer in the notes is still correct :-) )
G
grace3496·
For additive model, it should add up to 0 because the variations are above or less than 0?
J
John MoffatTutor·
Correct :-)
M
Martin·
will all of the above be required in the exam?
J
John MoffatTutor·
Not the whole exercise - just extracts from it (but you need to understand the whole exercise to be able to be able to attempt extracts).
You will find plenty of examples of how it can be asked in your Revision Kit (you must buy a Revision Kit from one of the ACCA approved publishers, because they have lots of exam-standard questions for you to practice).
R
rochwal·
Hi I think there is an error in the video, when calculating the Multiplicative Model quarterly average the Q3 amount was written as 91.35% and not 95.35%, just want to know if i miss understood.
J
John MoffatTutor·
Yes - you are correct. Sorry :-(
The answer at the back of the free lecture notes is correct :-)
S
Shaheen Fathima·
Have the videos been updated?? I see a few differences compared to the videos that I saw last week......
J
John MoffatTutor·
You can't have, because nothing has changed since last week :-)
S
Shaheen Fathima·
The reason I am asking is because the comments have disappeared from all the video lectures and there is a slight tuning prior to the point in the video where they say "This is a lecture from open tuition......." and also the headings to all chapters are printed in red (or blue I am guessing) and not written with the black marker like it was previously.....?
J
John MoffatTutor·
I don't understand. The lectures have not changed, and all the comments are still here :-)
S
Shaheen Fathima·
Why is it that the actual value (when calculating seasonal variation) is taken from the third quarter; why is it that the first two quarters of the year 2000 are not considered?
J
John MoffatTutor·
Because (as I do explain in the lecture) there is nothing to compare the first two (and the last two) quarters with.
What i wanted to find out myself.
Kindly assist sir
If anything is asked them most likely is just the reason for the two different ways of calculating the seasonal variations (i.e. whether or not to calculate them as %'s).
(The fact that you are using historical data is a limitation of any forecast, but it not a limitation of the method because any method of forecasting will be based on historical data.)
For the final question, you brought down 91.35 instead of 95.35 which changes the end answer. Where you got 398.26 I got 400.26 which led to a decrease by 0.065 and then a sum of 400. Is that correct?
So the total average would be 400.26 (like you said)
To make it 400%, we'd have to reduce the averages by a total of 0.065% (0.26/4)
So I reduced Quarter 1 and 2's averages by 0.06%, and Quarter 3 and 4's averages by 0.07%
Final figures: 99.95%, 102.96%, 95.61%, 101.48%
If the Question asks for the Average Seasonal Variances for the Quarters,
Should the answers be +0.06,+3.13,-3.94,+1.44 OR -0.1125,+2.9575,-4.1125,+1.2675?
Thanks!
I understand how we get -0.17, (0.69/4= 0.17) but I do not understand -0.18 in 24:16 minute.
Could you explain it, please?
Thank you for considering my request.
I want to know why the total avaerge must add up to 400 and 0 using the multiplicative and additive model respectively?
With the additive model that are all a bit more than zero or a bit less than zero, so the average is zero and the total should be zero.
Also when referring to the multiplicative model as an alternative to the addictive model, you made mention of a hypothetical sales increasing and it being ten times higher at the end as compared to the beginning, by that do you mean ten times sales increase when we compare the difference between the sales figure at the beginning of the tenth year as against the sales figure at the end of the tenth year? Or ten times sales increase when we compare the difference between the sales figure at the beginning of the first year as against the sales figure at the end of the tenth year.
Sorry for the long question I have been on this topic for some days now and I can't help but ask for clarity
If the sales are increasing then it would be reasonable to expect that the seasonal variation would increase as well. The additive model has the same seasonal variation whatever the underlying level of sales is. The multiplicative model means that with higher underlying sales we will adjust for a higher seasonal variation.
(Last trend - first trend) / (no. of set of data - 1)
I want to understand why we minus by 1.
Suppose there are four observations: 1, 2, 3 and 4
There are three differences 1 to 2; 2 to 3, and 3 to 4.
If using the results of example 1 to forecast using regression analysis, which sets of figures would be considered X and which would be Y?
Many thanks,
I don't understand why the centred average of 86 is said to be near quarter 3 than 2?
Please could you explain to me
Similarly the 4 quarters average of the next four quarters is in the middle - so is between quarter 3 and 4.
When we centre them we take the average of the middle of quarters 2 and 3 and the middle of quarters 3 and 4, and this then corresponds to quarter 3.
In the multiplicative model, the total averages arrives to 400.26%,so do we need to reduce the .26% to arrive at 400%??
At the moment they add up to a total of +0.69, so to make the total add up to zero we need to subtract 0.69. We subtract an equal amount from each of the 4 quarters. However because 0.69/4 = 0.1725 and we are only working to 2 decimal places, we subtract 0.18 from one of them and 0.17 from the others (and it does not matter at all which one you subtract 0.18 from).
?
However the idea should still be clear (and I think the printed answer in the notes is still correct :-) )
You will find plenty of examples of how it can be asked in your Revision Kit (you must buy a Revision Kit from one of the ACCA approved publishers, because they have lots of exam-standard questions for you to practice).
The answer at the back of the free lecture notes is correct :-)