Needing Direction: Showing Progress Over Time with Small Numbers

[SamSmith1]

Needing Direction
I have sets of number:
week 9 27 turned in 25 good 2 bad = 93% & 7%
week 10 56 turned in 55 good 1 bad = 98% & 2%
week 11 75 turned in 74 good 1 bad = 99% & 1%
week 12 6 turned in 5 good 1 bad = 83% & 17%

In a graph by % of total it looks like the bad went up drastically - however is was the same number "1"

How do I show a graph more meaningfully over time, when it's the total number that is affecting the % and not the result of 1

Should I look into ratio's of grand totals over time, or statistical significance, or difference from week to week, weighted numbers based on what? or is this the best that can be done?

 

[berkeman]

Welcome to PF.

Maybe try using a Bargraph plot by week #, with different colors for Good/Bad (maybe Green/Red?) and the height of each week's bar is the Total. That way the relationships will stand out better.

BTW, is this a schoolwork question or is it for your work?

 

[FactChecker]

Welcome to PF.

Maybe try using a Bargraph plot by week #, with different colors for Good/Bad (maybe Green/Red?) and the height of each week's bar is the Total. That way the relationships will stand out better.

That's a good idea. But in this case, will the bad 1 would even show up on an axis that goes up to 75?

 

[berkeman]

That's a good idea. But in this case, will the bad 1 would even show up on an axis that goes up to 75?

Excellent question. I considered posting the Excel plot, but if this is schoolwork, I'd have to give myself an infraction for "too much help". And I don't need any more of those!

 

[FactChecker]

Excellent question. I considered posting the Excel plot, but if this is schoolwork, I'd have to give myself an infraction for "too much help". And I don't need any more of those!

Good point. And I am not sure that there is anything to advise other than to try different plots and see what works best. Maybe two plots, one of the percentages and one of the numbers.

 

[berkeman]

Yeah, and what does "turned in" mean? In school, if you don't turn in your homework, that counts toward the "bad" number. Why is the number of things involved changing so much week-to-week.

We need the OP to respond before we waste more time on this question, IMO.

 

[Ibix]

Also, who's the audience? You might take different approaches with experienced statisticians from with management types.

 

[berkeman]

management types

 

[DaveC426913]

Maybe a logarythmic vertical scalebar graph? Very small numbers would be proportionately emphasized. Large, absolute values would be de-emphed.

 

[Ibix]

Some of my management is frighteningly competent at this kind of thing. Some of 'em, not so much...

 

[DaveC426913]

and what does "turned in" mean? In school, if you don't turn in your homework, that counts toward the "bad" number.

Only if its a mandatory assignment. If it were for bonus marks or some such then sample size over time is variable.

It's a good point you raise. If it is a voluntary assignment, then total is not the right context. If only six chose to hand in, that's not really important, and that 17% jump in "bad" is significant.

 

[berkeman]

I keep a "Double Secret Probation Penalty List" for students who don't turn in their "bonus" assignments. Just saying...

https://www.hollywoodreporter.com/l...t-now-john-belushi-tim-matheson-more-1126335/

 

[mfb]

Maybe a logarythmic vertical scalebar graph? Very small numbers would be proportionately emphasized. Large, absolute values would be de-emphed.

Then you can't stack them, but two bars next to each other (bad/good) would work.

If the audience has some experience with statistics then you can plot the percentage but indicate a 90% confidence interval in the graph. That will be very large for weeks with low totals.

 

[BWV]

I would use a stacked column, labeling numbers and %. Can add a bullet on the low number of events in the last week (perhaps something also that if the bad prob is 1% then there is around a 6% chance of getting at least one bad in 6 tries)

 

[SamSmith1]

Needing Direction
I have sets of number:
week 9 27 turned in 25 good 2 bad = 93% & 7%
week 10 56 turned in 55 good 1 bad = 98% & 2%
week 11 75 turned in 74 good 1 bad = 99% & 1%
week 12 6 turned in 5 good 1 bad = 83% & 17%

In a graph by % of total it looks like the bad went up drastically - however is was the same number "1"

How do I show a graph more meaningfully over time, when it's the total number that is affecting the % and not the result of 1

Should I look into ratio's of grand totals over time, or statistical significance, or difference from week to week, weighted numbers based on what? or is this the best that can be done?

Thanks everyone for your help. 4/4/22

 

[DaveC426913]

Thanks everyone for your help. 4/4/22

https://xkcd.com/979/

How many times have I Googled a programming problem, found someone else asked it, and then responded to their own question with "Thanks, I figured it out."

 

[berkeman]

Thanks everyone for your help. 4/4/22

So does that mean you found a solution? Can you post what you did?

And BTW, was this a problem for schoolwork?

And... why did you append the date to your latest reply?

 

[DaveC426913]

And... why did you append the date to your latest reply?

Indeed. This is the second reference I've seen to 4:4. No idea what the significance is but I think it's something woo.