AT Hiking Rates, Section by Section (AT Data and Schedules)

[map man]

I'm curious now that I've made several edits and additions to the main article over the last few weeks whether people think the article "works." Is it useful? Is the information presented in as clear a fashion as possible? Is it accessable -- that is, does it avoid being too complicated and avoid only appealling to statistics nerds like myself? I've arrived at the idea of referring readers to the article thread (specifically, post #28) for a more in-depth discussion of the data. Does that seem like a good idea? For those who've brought up issues concerning the statistical methods used, have your concerns been addressed in either the article or post #28? And finally, does the article in its present form meet the standards that an article here at WhiteBlaze ought to meet?

Now that the weather is nicer here in my neck of the woods (Iowa), I'm out hiking on weekends instead of messing around with the raw data used in the study like I was this winter, but I'd still like to hear from people if you have any suggestions at all to make the article better.

[sdoownek]

I always find it iteresting to see how mathematicians do things.

To that end, I note that there are several of you that have commented on this thread that seem rather, to use a term previously mentioned, "jerky". To those people, I would ask that you read the following commentary:

http://80below.com/archives/114-You-...ead-ones..html

While it doesn't directly apply to the discussion at hand, it does apply to the people willing to defame and unreasonably question the author.

Just a thought. Or a slight shift, if you will.

[joel137]

Just for amusement and comparison I thought I'd post a section hikers equivalent data

Winding Stair Gap -> Amicolola 10 days/115.4mi
Winding Stair Gap -> Hot Springs 14days/167.5mi
Hot Springs -> Pearisburg 26 days/343.5mi
Pearisburg -> Harpers Ferry 28 days/ 379.4mi
Harpers Ferry -> US 7 MA 35 days/ 499.4mi
US 7 MA -> Katahdin 51 days/ 646.1mi

Broken up in the same segments as listed for the thrus

Springer to GA border 8 days
GA border to Fontana 7 days
Fontana to Damascus 21 days
Damascus to Waynesboro 29 days
Waynesboro to Harpers Ferry 11 days
Harpers Ferry to DWG 19 days
DWG to Kent 12 days
Kent to Glencliff 22 days
Glencliff to Gorham 9 days
Gorham to Stratton 10 days
Stratton to Katahdin 14 days

The above is actual hiking days, I only did one zero day in damascus, so depending on how you want to count an extra day might be added there.

[roxy33x]

I am going to be doing an overnight hike with my husband and two dogs and I was wondering if anyone had any recommendations in the NC/TN area or anywhere near... I live in Charlotte. It just needs to be dog friendly... Preferebly on the AT.
Thanks

[cutman11]

Absolutely a great piece of work, map man. I am also a section hiker but have not completed the entire trail yet to add my section data, but I can say that the numbers generated correlate very well with my own hiking log as far as hiking days from GA to PA, where I ended last fall. I would be interested, if its not too much trouble to generate, a table of the location and frequency of the zero days, with some sort of "probability" that a zero would be taken at a certain zero day location, ie, Damascus- 95% probability of a zero...Hiawassee 45% probability of a zero,etc, that sort of thing. Or list the zero locations from most to least frequent. I have been keeping a notation of "zero days" I would have taken if I had been thru hiking, just to give myself a sense of the total days I would have used if I had done the hike all at once instead of sectioning. It would probably be useful for future thrus to consider "most likely" zero stops when planning their thrus.

[joel137]

One observation from my one data point

regarding differences between sectioning and thruing

It would appear that sectioners are less prone to taking zero days.

My guess would be that sectioners are more on a schedule than thru's.

I only did one, and in my current second go around there have been no zero days, I coubt there will be any more while sectioning, barring medical concerns.

With luck I'll get the chance to thru when I retire, in ten years with luck of a different sort. I imagine I will do zero days in the course of that journey.

[rickb]

Wow, that is great stuff.

To ask for more would be gready, right? But I will anyway.

I'd be interested to know how different demographics within your sample conform to the averages. Are there any differences between how men and women hike? Is age a signifcant factor of zero days? That sort of thing. Do people starting in early March end up spending much more time on the Trail than those starting in April? Are the fastest starters more likely to finish?

I guess what I am saying is that I really enjoyed what you did and hated to see the end of the thread!

[map man]

To respond to some recent posts:

joel137, my guess is you're right that section hikers take fewer zero days than thru-hikers, but I have no objective numbers to back that up. I'd say that your long section hikes (I know one was over 600 miles long) while taking no zero days at all (just that one in Damascus) are a little outside the norm, I would guess, even if section hikers in general do take fewer zero days.

rickboudrie, the demographic breakdowns you're talking about I can't really do right now with my data. I can't break the hikers down by age because not everyone at trailjournals.com reveals their age. (edit insert: And as of March 2008 there are only 26 female hikers in the study and that's too low still to include a gender breakdown in the main article) However, there is some info in post #28 in this thread about the relationship between start date and days to complete that you asked about. The group leaving in early March did take longer than those leaving in April or later, but again the problem is that once I start breaking the hikers into five or six groups based on anything the sample numbers start getting pretty small (I believe there were somewhere around 19 in that group that I'd define as "early March") to make reliable generalizations.

cutman 11, I think looking at the percentage of thru-hikers who take zero days in each trail town along the way, and comparing each town that way, would be a cool thing to know, and it had not occured to me until you mentioned it. The way I have the data recorded, though, I can't pull those numbers out right now (I could tell you how many zero days each hiker had in the DWG to Kent section, for example, but not how many for just DWG itself). I'd need to go back to each of the 173 journals in my study and figure it out. But when cold weather descends again next winter, and I hang up my hiking shoes for the year, and I have the time again to take on the task, that's something I will consider doing.

[cutman11]

thanks for the reply map man, and in a sense, the result of my request would be the defacto vote of which is the most popular "trail stop" !!!
Will look forward to your data when youre able to get to it.

[Vi+]

Map man,

Thank you for providing the results of your analysis.

You “... decided to (calculate) how long the typical (Thru-Hiker) takes to hike the AT (section by section).”

You gathered nominal data, crunched the numbers, and then reported some of your mathematical deductions as hundredths of a day. Some users have overlooked the humor inherent in appraising six months in 2.4 hour segments.

I never considered being able to determine exactly where I and everyone else would stand at every moment of every day. You’ve presented some very handy information.

Thank you, again.

[SGT Rock]

I'll get this one moved over someday as well :

[nhalbrook]

If there is info in the data to identify those who are seasoned hikers and those who are not, would be interesting to see what that separation shows in terms of zero and hiking days.

[Tramper Al]

Mapman,

This is great stuff - thank you for your efforts.

I too am well trained in and occupied in the practice of the methods that have been used and could be used to analyze such data. I am also well aware of the issues concerning sampling and generalizability here.

My only suggestion is with regard to "outliers". I think you should take note of your stongest, loudest, most zealous and pompous critic, and ignore him.

Thanks again, very interesting.

[Alligator]

...
I too am well trained in and occupied in the practice of the methods that have been used and could be used to analyze such data. I am also well aware of the issues concerning sampling and generalizability here.
...

It's OK to take the cheap shot Al, that one was pretty good. Whomever you are speaking to is probably laughing. If I were listening in to this discussion between well trained professionals though, I'd certainly want to know why you believe it's ok to gloss over the sampling and generalizability issues here.

I'd also be wondering Al, given your experience, what do you think about the mean differences in hiking times for the years 2001-2005 given in Table E, post #29. Like, do you think there are any significant differences between the years and how that might affect generalizations. I have some thoughts on the subject, but I'm a little bit shy so I'll let you go first. Let's hope that the zealous, pompous, LOUD, and STRONG critic doesn't pop in in the interim.

[Alligator]

What's up Doc, can't answer the questions? They're actually very relevant. Doubts are surfacing regarding your expertise .

[Alligator]

I didn’t want to leave this hanging. Initially, I raised the issue that year could be an important factor regarding hiking time. I was accused of being quite a number of things. Whatever. Here’s the crux of the problem that I see. If there are significant year to year variations, it could be very inappropriate to average across years. I have returned to this issue because the information regarding years was not initially included and was added later.

I took a moment to graph out the hiking day means by year. I put these into two figures. One figure uses a narrow range of days on the y-axis, the other I used a wide range to include the whole range of hiking days reported. The mean is plotted as a solid black line throughout. Looking at either figure, it appears that there are differences between years, most notably 2002 and 2005. There’s just about a 14 day difference between those two years, which is slightly less than 10% of the overall grand mean of 147.6 days. That’s a strong suggestion to me that there are temporal variations. Now, it could be that these two years are just random fluctuations. This would happen if the overall variance of the observations was wide. It’s hard to say without the entire data. If this had been a random sample, this could be tested with a simple ANOVA or a nonparametric test if the group variances were not equal. And if you wanted to discount the non-random selection procedure, it could be tested anyway.

Call me a stickler, but I’d want to know the whys of this potential difference before I started averaging over years.

[map man]

Some observations on how or if this article can be used for planning or gauging a thru-hike, now that I've had a few months to reflect on it:

First off, it's hard to know if the group I studied, thorough journal keepers (TJKs), are representative of all NOBO thru-hikers. You could make the case that maybe people so disciplined about journal keeping are disciplined in other ways too and more likely to get to Katahdin quicker than normal. Maybe all the time spent at computer terminals or typing away at pocket-mail devices would slow them down compared to other hikers. It's hard to know.

So I'd be hesitant to look at the "average" number of days in Table 1 in the article that it took TJKs to get to each landmark and conclude when comparing myself to this group that I was a fast or slow hiker. I wouldn't want hikers taking twenty days to get to Fontana instead of the 15.8 average days in Table 1 to get discouraged about their chances of completing their thru-hike. The same thing goes for those who happen to take more than the "average" of around twenty zero days on the way to Katahdin.

What interested me most when I set out to do this study was to figure out how hiking rates varied from one section of the trail to another. So I think it's more pertinent to know, for example, that the "average" NOBO TJK covers about 60% more miles per hiking day in the Damascus to Waynesboro section than in the Springer to Georgia border section. The exact numbers, sixteen miles per hiking day versus ten, aren't as useful for planning or gauging purposes as the underlying proportionality is.

And I can't see many reasons why this proportionality between sections would be significantly different for my study group, TJKs, than it would be for the entire NOBO thru-hiking population (though I can't prove that statistically). So what I think is the most useful and reliable are the numbers in Table 2 in the article, miles per day and miles per hiking day, when looked at for the underlying proportionality and not as exact predictive numbers. Likewise, the numbers in Table 4, where progress is projected for four, five, six and seven month hikes based on those proportions between sections, should also prove useful, I think.

Noting Alligator's reluctance to group hikers from different years and different starting times I decided to compare the progress of TJKs, breaking them into groups based on: 1) the year they hiked; 2) the starting date; 3) the speed of the hike (looking at the sixteen fastest hikers, for instance, who actually did average four months to complete the trail, etcetera). I computed the amount of time for the mean hiker to reach each landmark as a percentage of the total time to hike the whole trail. I computed the same thing for each of the sub-groups based on the three criteria I mentioned and I looked at the total days to hike as well as the "hiking days" (excluding zero days). So I ended up with six separate tables. I'm not going to reproduce all six here, but instead will show the one table where sub-groups varied the most from the overall mean percentages, to show just how closely even the most widely varying sub-groups compared to the overall mean.

The following table shows five different groups based on start date. I used the same grouping I did in Table C in post #28. The earliest group started Feb. 27 or before; group #2, Feb. 28-March 5; group #3, March 6-17; group #4, March 18-April 8; the latest group, April 9 or later:

(Table includes only hikers from years 2001-2005)


Table F -- Percentage of Total Hiking Time to Reach Each Landmark, by Start Date

Earliest~Group2~~Group3~~Group4~~Latest~~Mean~~Lan dmark
4.8%.......4.8%......4.9%.......4.7%......4.4%.... ..4.7%....Georgia border
9.4%.......9.3%......9.7%.......9.2%......8.7%.... ..9.3%....Fontana
24.3%.....24.4%....24.2%.....23.3%.....22.8%....23 .8%....Damascus
41.0%.....41.3%....41.5%.....40.8%.....38.8%...40.9%...Waynesboro
46.8%.....47.8%....48.8%.....47.6%.....45.6%....47 .6%....Harpers Ferry
60.2%.....58.8%....60.1%.....58.6%.....57.6%....59 .1%....DWG
67.1%.....66.5%....67.3%.....65.9%.....65.2%....66 .5%....Kent
80.6%.....80.1%....80.9%.....79.9%.....79.3%....80 .2%....Glencliff
86.1%.....86.2%....86.3%.....85.5%.....85.7%....86 .0%....Gorham
91.8%.....92.0%....92.0%.....91.5%.....92.0%....91 .8%....Stratton
100%......100%.....100%......100%......100%.....10 0%.....Katahdin


I put in bold the single number in all six comparisons that varied the most from the mean. TJKs leaving April 9 or later took 38.8% of their hike to get to Waynesboro while mean TJKs took 40.9% of their hike to get there. A very small difference. This late starting group took a little over five months to complete the trail and if you compute how many days the "typical" five month hiker takes to get to Waynesboro -- sometime on the 63rd day -- you find that a "typical" five month hiker leaving April 9 or later would get there between two and three days earlier. Again, a very small difference. And keep in mind, that is the single greatest variance of all the numbers I looked at for start year, start date, or hiking speed -- a lot of numbers.

So for the purpose of figuring out "typical" hiking progress based on proportions between sections, I think it's legitimate, and even desirable, to lump all 240 (the number for 2001-2010) TJKs together (and now my "Alligator-sense" is tingling -- I've got a feeling I'm going to hear from him on this one).

It was a useful exercise to go through, though. As small as the variances were, I did find out that start date made for a slightly greater variance than start year, and that both of these definitely showed more variance than that based on the speed of the hike. Not surprisingly, "hiking days" (excluding zero days) showed less variance from the means when comparing these sub-groups than total days did.

It would be better to be able to sample all AT hikers instead of relying on the self-reporting of hikers keeping journals at trailjournals.com, but how would one do it? Just imagine a modern-day Marlin Perkins and his buddy, Jim (for those of you old enough to remember "Wild Kingdom"), bringing down thru-hikers at Springer with tranquilizer guns and fitting them with radio collars or GPS devices and then watching the hikers wobble off down the trail as the tranquilizer wears off. Just picture it with me. Am I the only one here warped enough to find that entertaining?

[Mountain Dog]

Very interesting article and it can be used by a lot of us. Thanks for all of the good work.

[Alligator]

...
I'm not going to reproduce all six here, but instead will show the one table where sub-groups varied the most from the overall mean percentages, to show just how closely even the most widely varying sub-groups compared to the overall mean.
...

Theoretically, the usual accepted practice is to test whether the group means are equal. Then interest generally moves to which are different. Do the overall F-test, then use some multiple comparisons procedure to figure out which groups are different from which. This differs from what you have done because you have compared the groups to the mean. The differences are more pronounced when comparing the min proportion to the max.

Further, a more complex problem is that every section is correlated with the last, since the overall measurement is a proportion. I'm not trying to pull some mumbo jumbo here either, a multivariate stats class would help a lot. The proportions for each section are measured on the same hiker. There are some number of hikers in each group, with multiple measurements on each hiker. This creates a multivariate analysis. There would be a group mean vector. This group mean vector would have nine elements, representing the mean proportions for each section. The 10th section group mean proportion is redundant, because the proportions all sum to 1. Still with me ? Then, if one really wanted to know if they were different, one might use a multivariate test such as a multivariate analysis of variance MANOVA. Then univariate tests at for each section group mean. Technically, what you are interested in is if the profiles are nearly the same. See the chart in the attached spreadsheet. Most of these profiles do look the same, based on the scale of the graph. However, there is some difference between early and late starts, see the table. The differences are not the same for all sections. This means the profiles are not parallel, indicating a somewhat different pattern to the groups.

Other differences:
Group 3 took longer to reach HF then did early or late groups.
The latest group is taking the least amount of tim to reach Waynesboro, while the other groups are fairly close to one another. This suggests the late group is moving a bit faster, possibly feeling early pressure to move along or maybe better weather. I'm not sure.

I'm not bringing in the complex topic of MANOVA to cloud the discussion. What I want to point out is that the profiles could have been completely out of sync, like a sine and cosine wave, yet the manner in which you approached looking for differences, the univariate approach, could have completely missed a situation of that nature.

Normally, I'd test the data first, then look for the differences. It avoids data snooping . For this response variable, given the small differences, I think you have a pretty good case. That is the differences in proportions are small enough to not worry about the lump grouping.

BUT, you did change the response variable to a proportion in order to compare the groups. What you did was to in effect change the topic of discussion, answer the question as applied to a different measure, get the answer you wanted, then innocently presented this as a form of rebuttal to my original caution about grouping the TJK's . My argument centered on the total number of hiking days by years, and how the groups appear different. Lump grouping appears reasonable for one measure, but not the other. OK?

[Jack Tarlin]

I'm not remotely an expert on mathematics or statistics, so I'm not going to comment on Map Man's supposed methadology, biases, weaknesses, etc.

What should be mentioned, tho, is that in my opinion, the information he's providing re. how long it takes the average person to hike a certain section, is, in my opinion, quite accurate information indeed. There are perhaps one or two places where I think the "typical" thru-hiker (if indeed such a thing exists) might in fact require a bit more time than MM's figures indicate, but on the whole, these are very minor quibbles. However he obtained or came by these figures, I happen to think they're pretty accurate, and therefore, I think this article can be very useful for folks in the planning/preparation stage of their trips, especially those who are attempting to figure out their approximate itinerary, schedule, etc. This is particularly useful information for folks who are wondering if they're giving themselves enough time for their hike; or folks who need to be in a specific place at a specific date and are wondering how much time they'll need to get there, etc.

I think this is a great piece of work and a very worthy contribution to Whiteblaze.