Comments on: The Great Input Showdown!

Yeah, a good follow-up review might be various software alternatives like Fitaly, MessagEaseST, TapPad, ....

As with my other reviews, please feel free to post direct questions in the comments section and I'll check back and answer them the best I can.

Cheers,

Dan

I agree that the Fitaly keyboard's a good layout; when I had a Palm IIIc, I wanted to use Fitaly, but it was incompatible with color devices (at least it was at the time). Luckily, Gustavo Broos wrote a keyboard hack that allowed you to make your own keyboard layouts (I believe the program's called Virtual Keyboard), and I wrote a Fitaly-like keyboard for it, and it worked wonders. Yes, it's an on-screen keyboard, but it worked great. If you're interested, I could put it out on the Net somewhere...

I haven't used FITALY myself, but I believe it is a good thing. My question is why PalmSource, Plam or Sony can not implement both Grafitti and something like FITALY in their products. This will emilinate the need for thumboard, while still statify those Grafitti zealots.

It is just a remap of Grafitti or VG area.

Dan,
My apologies for hijacking your review to talk so much about FITALY, but I just love FITALY and software keyboard. I was greatly honored to be on the BETA testing team and used FITALY for 8 months on my miserable mono Ipaq.

I might note that FITALY final build is due out on November 1 and it is really worth getting.

i have to agree on this, because like me, i get 90 WPM on the full keyboard, about 39 with the mini keyboard of my NR, and about 30 with grafitti... so everyone varies, from if you are a long time grafitti user, to a new biginner like someone who would be buying the zire.

____________________
"WE DONT WANT YOUR STD's!!"
www.pocketloft.com

I think I know what you mean, so here is my take.

When I switched from my IPaq, using onscreen keyboard/Fitaly to the Nokia 9210 I found I could enter data much faster.

The 9210 has a built in qwerty keyboard. The keys are rubber and it is not like a full Psion keyboard but OK.

If I actually timed how long it took to type a specific sentence then then I found that there was not that much differece in speed (for short entries).

However, I still seemed to find using the keyboard better. I finally figured out what the reason was.

When using the on screen entry methods, I had to think about the data entry.

When using the Nokia, I could think about what I was writing.

In other words, using a "real" keyboard allowed me to focus on what I was writing rather than and inputing the text.

This is robably due to having used qwerty keyboards for many years, but I suspect it applies to most people.

Zuber

Fitaly is quick, like you said takes time to get upto speed though. I can't see ever reaching the same performance level with Graffiti.

Zuber

I just purchased the same keyboard myself. Fantastic design, works great. Only one downfall though and it's that pda holder / mirror thing they package with it. Actually, I never intended to use the mirror / holder because the software has the option to let you simply "flip" your screen. That way you can butt your pda right up against it and work on a flat surface.

Sounded great, but one more problem... The software allows you to "flip" the screen on all PDA's except Sony Clie models... Why? Good question. It's not even mentioned on the packaging. The only response I got from the manufacturer was that they would look into it. It took them a week to reply to my question as well.

Still, all that complaining out of the way, I have to say I really do like it. It beats carrying around the old bulky folding keyboard and I'm using it daily, even though I have to look at my clie upside down. Strange thing, is that I'm getting used to it.

Blessed are they who don't,
for they shall not,
nor will they.

Thanks Scott - you make a number of good points, particularly about having the stylus available for Graffiti alpha-numeric input as well as tapping on-screen UI elements.

pontif, my experience agrees very strongly with yours. Let's hope that Sony will work on better drivers!

One limitation that you didn't mention is the "Alt" key, which allows use of accented characters etc. The characters aren't marked anywhere on the keypad, nor could I find any reference to how they were mapped other than in the paper instruction book. Do you know of any reference that you can view on the Clie itself?

In terms of the connector cable, I think that design decision was made for compatability across different Clie handheld form-factors. As you stated, it is certainly less convenient to put on / take off than some other ThumbBoard designs I've seen.

- Dan

LOL - hey ska, you're not Charlie Krebs (my old stats prof from University days) are you? ;-)

Strictly speaking you're right on the money, but I felt discarding the lowest 5 was justifiable if you take into account the "fumble-finger" effect. e.g. I'm not used to the narrow-diameter stylus that comes with the new Clies and a few times my grip slipped too much and had to adjust (losing valuable seconds). Or I hadn't gotten *really* familiar with the thumbboard and suffered brain-lockup a couple of times. Or my 7-year old came down and distracted me, etc. (Yeah, excuses, excuses...)

Anyway, I hope you (and Charlie) will excuse the quasi-scientific statistical methodology :) I should have also used a much larger sample size but there's only so much time in the day, and frankly, I'm getting sick the "The quick brown fox..." ;-)

Cheers,

Dan

The grafitti and mini keyboard number is pretty close, the discarded data might change the average so close that it reach 1s. A difference that might means to somebody between a decission of splurging money on expensive mini keyboard to improve his work, or..

buying into a total gimmick.

plus the entire point of having standard deviation is to tell the consistancy of your typing between data take, hence the overall quality of data. that std above are just a cute numerology.

Geez, a lot of armchair statisticians on here. I'm a real statistician, and I think his method was just fine. He's not trying to generalize to all tests, he's trying to generalize to all valid tests. He applied his criteria in a consistent manner across all the entry methods. You can argue about his definition of a valid test and whether he should have thrown out 5 or 3 or 10, but he stated how he defined it clearly, so we can interpret his numbers in that context. Similar techniques are used everyday in lab studies.

He stated himself that the stylus behaves inconsistantly between test. Hence it is quiet possible that the bottom five numbers sets are containing two different noise which contribute two totally different values between grafitti and minikeyboard.

Hence, the 2 sets of 5 values might move the average and std value in two different size and direction.

kramsauer, he did have evidence that they belonged to a different population. He felt his hand slip on the stylus and he experienced his 7-year old interrupting him. To eliminate bias, he used a systematic approach to eliminating this noise, which is perfectly fine. Keep in mind the point of his numbers was for comparison between input methods, not to interpret the actual numbers for each input method. So his WPM numbers for each input method are inflated, but the comparison between them is much cleaner than if he hadn't used this approach, since his interruption rate between the methods was probably different.

I guess this shows that the signal to noise ratio isn't much improved by requiring a login...

(At this point there are more posts about the stats methods than about the keyboards or their usage, pros/cons, etc.)

Please, no flames. I just see the stats error as minor and only marginally relevant to the actual reviews of the keyboards.

If you really want valid stats, get 100 or more people in the same room for a week, keying the same thing over and over until they are totally proficient with it, then take the timings. Repeat that process with the same group and each keyboard. Then you have a sufficient sample to argue over the actual calculation method... Maybe.

I agree, this is very off-topic, but I'll make one last post to answer ska's questions. Earlier I used the term "valid" observations, but the more I think about it, it probably should have been "ideal" observations. It seems like the reviewer's intent was to determine differences in entry speed under ideal conditions for each of the methods, so he attempted to eliminate observations where he was distracted or just "messed up" (stylus slipping, fingers slipping on the keyboard, etc) without bias by removing the 5 lowest values from each of the input methods. I'm not suggesting that he got distracted the same number of times for each method, but under the null hypothesis of no difference between the methods, if you eliminate the same number of valid observations from the low end of the distribution, then the comparison will still show no difference, so that's why this is unbiased. Your point about stylus slipping is a good one if you were trying to get an accurate measurement of stylus entry, but this is why I think he was trying to get the measurements under "ideal" circumstances and "valid" was a poor choice of words.

It's not like he invented this method. There is a common statistic called the trim mean that uses this exact method, and is designed for this type of purpose. Usually this statistic would trim observations off the top and bottom, so the ensuing mean wouldn't be inflated, and if you want to criticize him for anything statistical, it should be that. But I still think his numbers are fine for comparison purposes, and I'm surprised his statistical methods have drawn so much scrutiny.

That's a great point. I can't believe how fast I got used to the keyboard on my NR, and I make fewer errors than with graffiti, plus I'm about 50% faster on the keyboard. Even so, what puts it over the top is that most of my input is with full-screen apps, which the keyboard is able to take full advantage of.