Grede inflation

I agree with you. In the publication where I worked, we didn't add 'stars' or a one word sum up like here at DPR since we never wrote up anything we'd not 'recommend'.
--

-----
-paul

Marcamera said:

"GREDE INFLATION"

50% words mis-spelled.

Rgds

Click to expand...

Thank-you. I am indeed an awful speller (and congenitally unable to proof read).

However, to quote Brian A, "Your statistics are totally invalid". After all, of the 14 letters I used, only one was incorrect, a 7% error rate. (Just kidding - you were quite right).

Regards

Simon

Simon Rodan said:

Marcamera said:

"GREDE INFLATION"

50% words mis-spelled.

Rgds

Click to expand...

Thank-you. I am indeed an awful speller (and congenitally unable to proof read).

However, to quote Brian A, "Your statistics are totally invalid". After all, of the 14 letters I used, only one was incorrect, a 7% error rate. (Just kidding - you were quite right).

Click to expand...

Tks for taking it in jest. I was not seriously criticizing the error, just trying to squeeze some humor in. Did not quite work.

And , actually, only 13 letters were used. One , "E" was used twice.

Rgds

Simon Rodan said:

Simon

Click to expand...

Member said:

Tks for taking it in jest .. ..., just trying to squeeze some humor in. Did not quite work.

Click to expand...

On the contrary - I think it worked very well. No offense taken

Member said:

And , actually, only 13 letters were used. One , "E" was used twice.

Click to expand...

Perhaps Brian A is right. I should learn to count before taking on more difficult problems that involve averages and other complicated statistical constructions.

Simon

Simon Rodan said:

Simon Rodan said:

Tks for taking it in jest .. ..., just trying to squeeze some humor in. Did not quite work.

Click to expand...

On the contrary - I think it worked very well. No offense taken

Simon Rodan said:

And , actually, only 13 letters were used. One , "E" was used twice.

Click to expand...

Perhaps Brian A is right. I should learn to count before taking on more difficult problems that involve averages and other complicated statistical constructions.

Click to expand...

Now, if you two don't settle down, you'll BOTH be sitting on the naughty step... [tsk!]
--
Regards,
Baz

I am 'Looking for Henry Lee ' (could be Lea, or even Leigh) and despite going 'Hey round the corner', and looking 'behind the bush', I have not yet found him. If he survives, Henry is in his mid-60s, British, and quite intellectual.

What is it all about? Well, something relating to a conversation we had in the pub 35 years ago has come to spectacular fruition, and I'd like him to know how right he was.

If you know somebody who could be this man, please put him in touch with me. Thank you.

Barrie Davis said:

Simon Rodan said:

Barrie Davis said:

Tks for taking it in jest .. ..., just trying to squeeze some humor in. Did not quite work.

Click to expand...

On the contrary - I think it worked very well. No offense taken

Barrie Davis said:

And , actually, only 13 letters were used. One , "E" was used twice.

Click to expand...

Perhaps Brian A is right. I should learn to count before taking on more difficult problems that involve averages and other complicated statistical constructions.

Click to expand...

Now, if you two don't settle down, you'll BOTH be sitting on the naughty step... [tsk!]
--

Click to expand...

I'd give anything for that. You mean getting young again?
Rgds

The problem is that you are taking a discrete ordinal variable and assigning integer values, which is fine up to a point. From that you could do min, max, and mode. But to calculate an average means you are assuming more than was in the original data. You are assuming the existence of a zero, the position of the zero is defined, that there is a constant interval between the values, and in this case, that the interval is the same between zero and the first value as between the others.

For example, instead of assigning 1-5 to the categories, it would be equally valid to assign values of 96, 97, 98, 99, 100; or -2, -1, 0, 1, and 2, or even 20, 40, 42, 67, and 78. The mean for these assignments would be very different from the one you calculate.

Consider another nominal categorical variable, the Saffir-Simpson scale for hurricane categories, which classifies hurricanes into five categories, mostly based on wind speed. In this case it would be false to assume that a Category Two hurricane had twice the wind speed of a Category One hurricane, the intervals aren’t constant – category One = 64-82 knots and Category Two = 96-113 knots.

Brian A.

Hugowolf said:

The problem is that you are taking a discrete ordinal variable and assigning integer values, which is fine up to a point. From that you could do min, max, and mode. But to calculate an average means you are assuming more than was in the original data. You are assuming the existence of a zero, the position of the zero is defined, that there is a constant interval between the values, and in this case, that the interval is the same between zero and the first value as between the others.

For example, instead of assigning 1-5 to the categories, it would be equally valid to assign values of 96, 97, 98, 99, 100; or -2, -1, 0, 1, and 2, or even 20, 40, 42, 67, and 78. The mean for these assignments would be very different from the one you calculate.

Consider another nominal categorical variable, the Saffir-Simpson scale for hurricane categories, which classifies hurricanes into five categories, mostly based on wind speed. In this case it would be false to assume that a Category Two hurricane had twice the wind speed of a Category One hurricane, the intervals aren’t constant – category One = 64-82 knots and Category Two = 96-113 knots.

Brian A.

Click to expand...

Finally something makes sense.
I am buying category One = 64-82 knots cameras from now on.
Rgds

My main consideration is - will it do the job I buy it for?

Will it take a shot of a radio amateur installing an antenna before he has come down the ladder again? (My super-zoom did while my mate on his L_ a was still fiddling with the aperture and focussing long after the guy had finished his job and was at the sausage sizzle by the time Alan had finally got his camera set up, looking at a ladder with nobody on it.)

I could go on and on. But almost all the cameras which do the job I bought them for never even got a review here.

Sandra's daughter has brought with her a guy called Scott, who is a camera salesman. He makes the same observation: A number of camera models that get rubbished in reviews are perfectly suited for the tasks that his customers have bought them for.

Two months ago I bought a 7x zoom pocket camera (Olympus Stylus 7020) without reading any reviews that could guide me.

My reasons were as follows:

It does not have the mode dial right next to the shutter button.

It does not have a pop-up flash (remember, it is a camera for storing in a pocket).

It has a new lens with more aspherical elements, expected to remain sharp through a 7x zoom range, where 10x or 12x zoom could prove iffy.

I have good experience with the movie sound from other models of the brand.

The Stylus 7020 meets my expectations. The battery also lasts longer than I dare hope for from the 750 milliAmpHour rating.

Henry

--



Henry Falkner - SP-570UZ, Stylus 7020, Stylus 800
http://www.pbase.com/hfalkner

Henry Falkner said:

He makes the same observation: A number of camera models that get rubbished in reviews are perfectly suited for the tasks that his customers have bought them for.

Click to expand...

Hmm, what camera has been "rubbished" in a review? Isn't the complaint in this thread that they are all at least "above average"? There seems to be a contradiction here -- unless you actually meant cameras are "rubbished in discussions of reviews on the forums..."

--
Erik

Hugowolf said:

The problem is that you are taking a discrete ordinal variable and assigning integer values, which is fine up to a point. From that you could do min, max, and mode. But to calculate an average means you are assuming more than was in the original data. You are assuming the existence of a zero, the position of the zero is defined, that there is a constant interval between the values, and in this case, that the interval is the same between zero and the first value as between the others.

For example, instead of assigning 1-5 to the categories, it would be equally valid to assign values of 96, 97, 98, 99, 100; or -2, -1, 0, 1, and 2, or even 20, 40, 42, 67, and 78. The mean for these assignments would be very different from the one you calculate.

Consider another nominal categorical variable, the Saffir-Simpson scale for hurricane categories, which classifies hurricanes into five categories, mostly based on wind speed. In this case it would be false to assume that a Category Two hurricane had twice the wind speed of a Category One hurricane, the intervals aren’t constant – category One = 64-82 knots and Category Two = 96-113 knots.

Brian A.

Click to expand...

Brain

Thanks for the explanation. Here's why I don't think it matters.

First, the value of the mean per se doesn't matter. I'm looking at changes relative to the mean and whether the mean was 0 or 100 changes relative to that point are still the same.

Second, the hurricane scale is different in that it starts from a continuous variable (wind-speed), translates that into into categories - it's worth noting here that the mapping is non-linear - so then translating back is 1) inaccurate because information has been lost in translation and 2) the back-translation is linear while the first is not. However, in the camera rating scheme I am starting from a categorical variable, not one that is derived from a real ordinal scale.

Now clearly this is problematic if one were calculating regression coefficients since the assigned numeric values are arbitrary. Is Highly Recommenced 20% better than Recommenced (which would be the case for my assignments) or 1% better (if I had I started the scale at 95)?

Is the scale linear? I don't know, but it's an approximation we make all the time.

For example, we regularly use Likert scale and treat what is strictly speaking a categorical variable as cardinal integers. For example we often code Likert scales like this: "Very dissatisfied" = 1, "Dissatisfied" = 2, "Neither satisfied nor dissatisfied" = 3, "Satisfied" = 4, "Very satisfied" = 5; similarly for that almost ubiquitous "Strongly disagree" to "Strongly Agree" scale.

My point here is that we regularly make this kind of linear assignment of cardinal numbers to categories of an affective state. The sin I commit here is far less egregious than the Likert scale uses I note here, since the Likert assignments ARE usually used in estimations of regression coefficient.

Moreover, none of this matters when I am counting categories. When I note that there are, for example, 70 reviews in the top two categories out of a total of 100 conducted in one year. and observe that over the 10 year period 500 have have been in these two categories, I am not assigning any value to the category, only counting occurrences.

When I noted "The proportion of recommended and highly recommended were above the mean (79%) in every year since 2004 and only once before then." All I was saying is that the proportion of reviews in the top two categories each year, has been higher than the proportion of top-two category reviews in all years,

every year since 2004 and only once before then - but this does not involve the assignment of cardinal values, so the issue is not here.

I didn't explain my methodology clearly enough, but then this was a posting not a paper, and I was trying to be brief.

Hope that helps.

I fear, however, that it's not just the "overall rating" that is suffering

Consider these two images:

http://j.mp/5mt83j
http://j.mp/4RLPWo

To my eye, the lower image shows much more detail - the red and the yellow woven straps have more textrure in the lower image than in the upper one, and the black strap has texture in the lower image which is completely missing from the upper image.

The lower image is the camer JPEG and the upper one the Adobe ACR conversion. Yet, Andy Westlake writes. "Even at default settings with no real tweaking, though, ACR is producing a much more appealing and detailed output than the JPEGs - the X1 is clearly a camera that benefits greatly from shooting in raw". Maybe it's just me but Mr Westlake's description seems to me completely at odds with the evidence from the images.

Simon Rodan said:

To my eye, the lower image shows much more detail - the red and the yellow woven straps have more textrure in the lower image than in the upper one, and the black strap has texture in the lower image which is completely missing from the upper image.

Click to expand...

Aggressive sharpening sometimes works to improve detail and sometimes just makes it up. Look at the textures on the faces of the bill engravings - the X1 JPEG aliases the detail into slanting in the entirely wrong direction.

Simon Rodan said:

Maybe it's just me but Mr Westlake's description seems to me completely at odds with the evidence from the images.

Click to expand...

It's also a matter of preference: if you like highly sharpened images and don't mind the accompanying artifacts, then you may see it that way. Image qualities - particularly the weighting of different factors - is always subjective.

--
Erik