Can you point to both studies? I've only found multiple references to one study.
This website describes a simple regression (with 4 to 9 degrees of freedom depending on your underlying assumptions, let us assume 9 which is the "worst") that has a good match to that data of 38 points; if you are familiar with the quality of nutrition "science", this is actually extremely good match - e.g. atwater factors are often 30%-50% different than expected (and assumed "ground truth" based on 100 year old measurements) when actually measured. You have to not trust either atwater or the more recent measurements - possibly both.
Now, this regression, seems to hold very well with the exception of potatoes. From this, personally, I would infer either (a) a mistake somewhere in the data collection or (b) potatoes are special among the foods tested in a way not captured in this regression.
Actually working in data science for a living for a few years, my experience is that the practical thing to do is to drop the outliers from your dataset on one hand (so that your regressions and descriptions are robust with respect to the data that you DO consider) and investigate the outliers on the other hand, to figure out if it's an error in the input or a missing input (e.g., I would guess measurements like "water per cm^3" and "weight per cm^3" might be good measurements - if I had a proper "test set" to validate them against; just adding more measurements is usually NOT what you want to do).
Potato could be special in its effects on hunger / satiety - I don't know. e.g. among common food items, fresh eggs and fresh milk are special in the sense that, unlike any other "single" food item, they are each individually enough to sustain life. But based on the data I can find by googling (hardly anything to base policy on, but what I personally used as input) - I assume that, for whatever reason, potatoes are an exception to the general rule.
You may disagree, but I hope you at least understand my response.
>Actually working in data science for a living for a few years, my experience is that the practical thing to do is to drop the outliers from your dataset
The website's numbers are not a dataset. Again, it is made up numbers by a random website. There is no evidence behind it at all. The study's numbers do not have outliers, you only want there to be an outlier by comparing the dataset to completely made up numbers with no basis in fact.
> completely made up numbers with no basis in fact.
Did we look at the same website? They give the formula by which they derived the numbers, based on "standard" nutritional data for those foods. It's not a dataset per-se, it is (supposedly) a simple and easily reproducible computation of a widely available dataset. (No, I did not try to reproduce though I would if I were going to take any action based on this).
Why would you assume it is random?
And I ask, again, if you can point me to the two independent studies that confirm potato's properties, because I can only find one.
Suit yourself. The "calorie count" is just as invented according to your definition -- if you are not familiar with how calories are measured, educate yourself (and if you believe it is science, then make sure you can answer the question "then how come gasoline (12kc/gr) or paper (a carbohydrate) are not human edible", and "when and how were atwater factors derived". If you can answer these things and believe anything in nutrition is not "invented science" by your definition, then we have differing definitions of what constitutes science.
And if you believe (as I do) that all of this is entirely unscientific, but some of it is sort of useful - than what exactly are you arguing?
This website describes a simple regression (with 4 to 9 degrees of freedom depending on your underlying assumptions, let us assume 9 which is the "worst") that has a good match to that data of 38 points; if you are familiar with the quality of nutrition "science", this is actually extremely good match - e.g. atwater factors are often 30%-50% different than expected (and assumed "ground truth" based on 100 year old measurements) when actually measured. You have to not trust either atwater or the more recent measurements - possibly both.
Now, this regression, seems to hold very well with the exception of potatoes. From this, personally, I would infer either (a) a mistake somewhere in the data collection or (b) potatoes are special among the foods tested in a way not captured in this regression.
Actually working in data science for a living for a few years, my experience is that the practical thing to do is to drop the outliers from your dataset on one hand (so that your regressions and descriptions are robust with respect to the data that you DO consider) and investigate the outliers on the other hand, to figure out if it's an error in the input or a missing input (e.g., I would guess measurements like "water per cm^3" and "weight per cm^3" might be good measurements - if I had a proper "test set" to validate them against; just adding more measurements is usually NOT what you want to do).
Potato could be special in its effects on hunger / satiety - I don't know. e.g. among common food items, fresh eggs and fresh milk are special in the sense that, unlike any other "single" food item, they are each individually enough to sustain life. But based on the data I can find by googling (hardly anything to base policy on, but what I personally used as input) - I assume that, for whatever reason, potatoes are an exception to the general rule.
You may disagree, but I hope you at least understand my response.