Graph 3: Plotting of Linear Equation
Domain: 0 t 50
Range: 185 h 235
Let us now construct a table of values from the linear equation (y = 1.02x + 187). A data table was constructed using a calculator, and can be seen below in the Appendix (Figures 2 and 3).
Upon converting that data to a table and simplifying it to three significant figures; we arrive at this table.
Information Table 2 (Table of values from linear equation y = 1.02x + 187)
* - Although there are values for 1940 and 1944 (8 and 12 years elapsed respectively); as the original data set does not have such data for those years, we reject the two h-values.
Now we may compare the data points with the linear equation with the original data (reproduced below) to see what points nearly intersect each other.
Information Table 1 (Original data set)
Looking at these value tables and the graph above (Graph 3); we see that while the function approximately interests with four data points; at the 24th, 28th, 36th, and the 48th year-elapsed point; with approximately the same value. The function deviates from the original data set at the 4th year-elapsed point, with a difference of 12 centimeters. While the linear function meets at the last data point at 48 years with 236 cm; it also has a wide difference at the starting point – the very first year with a difference of 10cm. The linear model also faces issues that it does not account for a sinusoidal deviation from the equation line; for example, the 4th year and the 16th year had very different ranges from the function; and were extreme in range from each other; as the linear line goes in a straight fashion, it cannot account for these outlier points. Another weakness in the model is that the function will continue to increase over value in time; however this is not possible in this real life situation as there will be a limit to how high a human can jump due to gravitational forces; which the linear function cannot account for. On the other hand; the equation keeps decreasing in value when going back; and does not account for that time cannot be negative; nor that a high jump cannot go “down” below the ground (negative value) or remain on the ground (zero). However, for a linear model; this equation appears to be the best fit possible; as a rather precise method - Least Squares Approach, which involved all data points, was used rather than conducting a simple slope formula of two points to discover a gradient.
As discussed above, while a linear function works well as a analytical approach; using technology to determine a best-fit would be better approach to discover a better fit than the original linear equation. Let us explore an alternative method of finding a function; this time using regression. Using the Autograph programme’s best-fit algorithm enables the use of technology to construe a function of best fit using the original data table (below).
Information Table 1 (Original data set)
Autograph’s suggested best-fit function:
Let us take a look at this quintic function and how well it matches with the data set.
Graph 4: Plotting of Quintic Function and Linear Function
Domain: -2.5 t 50; Range: 150 h 290
Let us now plot the table values for the quintic function so we can compare the values between all three tables. Using the Autograph table plotting apparatus; a table of data has been generated for the quintic function (see Appendix, figure 3). After simplifying it to three significant figures, we get:
Information Table 3 (Table of values from: )
* - Although there are values for 1940 and 1944 (8 and 12 years elapsed respectively); as the original data set does not have such data for those years, we reject the two h-values.
Let us now compare the table of values to the other tables wrought from the original function and the linear function (reproduced below):
Information Table 2 (Table of values from linear equation y = 1.02x + 187)
* - Although there are values for 1940 and 1944 (8 and 12 years elapsed respectively); as the original data set does not have such data for those years, we reject the two h-values.
Information Table 1 (Original data set)
We can determine from Graph 4 (above) and the set of values that the quintic function approximately meets with many points of the original data set. The starting point; the 4th year, the 20th year, the 40th year; and the 48th year have the same values in height; with the 16th year, the 28th year, the 44th year are only one cm apart from each other. This demonstrates that the modeling of a quartic graph to represent the original data set is superior to a linear function, as the quartic graph shares many more intersecting points with the original data than the linear function. This is to be expected, as a quartic line has involves many turning points and inflections in comparison to a linear line which has a straight gradient. Note that the quintic line goes in a sinusoidal motion, weaving throughout the points, while the linear line generally stays in the middle of the points due to its nature. Interestingly, the quintic function shares some points with the linear function; at the 32nd year and the 48th year. All three functions share the final data point; the 48th year elapsed.
The nature of this data set raises a hypothetical question; what would happen if the Men’s High Jump Event of the Olympics occurred in 1940 and 1944? What would the gold medal height be? Even though that it actually didn’t happen; using the functions we have already derived from other data, such as the quintic and linear function, we can make an educated guess. Let’s take a closer look at the graph during the 1940 and 1944 time period.
Graph 5: Graph during the 1940 decade
Hered
We can see from the graph that between 1936 and 1948; the two different functions engage in different behaviour. The linear graph remains increasing in value due to its gradient; and the quintic graph decreases in value but reaches a turning point at approximately 1946 and starts to increase in value again; demonstrating sinusoidal behaviour.
Let us turn our attention to 1940. The quintic function’s value at this year is roughly 200 centimeters, while the linear function shows roughly about 195 centimeters. In the hypothetical case that the Men’s High Jump occurred on this year the record would presumably be approximately 200 centimeters. The rationale behind this statement is that primarily the quintic function gives a value of 200 in 1940; which seems much more accurate than a value of 195 centimeters from the linear graph; and that during a war, new, higher records would not likely be set due to bigger priorities such as war-making; hence the reduction in height for the record. In accordance with the above rationale; it would be likely that if an event were held in 1944 as well; the gold medal height would be around 198 centimeters because of athletic events being a low priority in the face of war. This would also be in tune with the upward curve from 1948, as that signifies that in the post-war era, sports and athletics becomes a priority again for the respective governments.
Trends within functions can also be used to predict future values to a limited extent. Let us attempt to discover what the height will be for the record in 1984; assuming that the functions remains true. Take a look at the graph below.
Graph 6: Functions showing future heights based upon trends: 1984
However, there are flaws in this hypothetical argument. The hypothetical height records are based upon the fact that there was a war going on; that athletes were not being competitive due to bigger pressures from the government to pay attention to the war, and not a superficial sporting event. If there was no war, then this scenario of a reduced value would likely not have existed; it would be likely that the value for 1940 and onwards would keep increasing; and not go down in value until a low in 1948. Hence, this hypothetical case has some weaknesses.
According to the quintic function; the gold medal height in 1984 will be approximately 251 centimeters. On the other hand, the linear function gives a value of 239 centimeters for 1984. It appears that in this case, the linear function seems to be more realistic in this aspect; giving a value of 239 centimeters. The quartic function value of 251 seems very high due to the exponential growth exhibited from 1976; exponentially increasing in slope as the value of t increases.
Now let’s have a look at what would happen in 2016; assuming that the functions remain true. Take a look at the graph below.
Graph 7: Functions showing future heights based upon trends: 2016
If we take a look at 2016, we see that the linear function would give a value of 273, meaning that the record for that year would be 273 centimeters high. The quartic equation however gives a much higher value of 330 centimeters; 3 metres and 30 centimeters high for the gold medal. Again, it seems based on past trends very unlikely that humans are able to break a world record of 330 centimeters, let alone 273 centimeters by 2016. This is an example of where regression is a very imprecise art; if you take a look at the quintic graph (see Appendix, Figure 4) the values increase exponentially up to 2008 the record then being 600 centimeters; then sharply drop exponentially; 2016 being a passing point when the values are dropping exponentially. The line continues then below zero onto negative infinity. Hence, as humans can’t really jump 600 centimeters nor does jumping in a negative length seem valid; it seems evident that a quartic graph, while very useful when within the range of data, becomes rather unrealistic when out of the range of data values. The same goes for the linear function; which steadily increases as each year passes by; if this were true one day humans could set a record jumping 100 kilometers up in the air; which is blatantly against the law of physics and motion. Hence, the winning height in 2016 is rather unrealistic given the current situation.
Let’s now expand our original data set from 1896 up until 2008. Shown below is the old data merged with the new information we have:
Expanded Given Information Table
Let us now plot the additional information we are given and see how it fits.
Graph 8: Extended Data Graph
It is immediately apparent that the quintic function becomes extremely inaccurate when expanding the data set. While the quintic function remained very accurate within the original data range of 1932 to 1980, once outside the ranges the function went to extreme values which would be essentially impossible to achieve given gravity, friction, and the fundamental law of physics; for example jumping a negative height or 6 meters high in the air – the extreme example being the time period between 1984 and 2020; where the values of the quintic function go to extreme ranges compared to the actual plotted data; rendering the quintic graph extremely inaccurate. The same applies to the time period before 1932; where the quintic graph originates from negative infinity; again very inaccurate compared to the actual plotted points in the pre-1932 era.
On the other hand; the linear function does not seem to deviate to the extremes that the quintic graph does; while it does not portray the data to the extent that it can be deemed accurate; it follows the general trend that the data is undergoing; positive increase. However, the linear function itself also is not without its weakness that it keeps linearly increasing and does not decrease its gradient; going up towards positive infinity; meaning that one-day should humans could be able to jump over 100 meters high should the trend be true. Yet, it should be noted that the linear function really did follow the real data points throughout the whole table; even in the pre-1932 era.
It would be apparent that of the two, the linear function better fits the additional data given.
Let us now take a look at the expanded data points by themselves.
Graph 9: Raw data plots
The overall trend from the graph (previous page) appears to demonstrate a positive increase throughout the passing of time; although there are parts where the values decrease; in generality, the values keep increasing. This is in line with real life as athletes compete not only to win but to break the previous records of other athletes. The growth appears to have the general shape of a logarithm; increasing but over time increasing less and less. This is also congruent with real life as there will be a natural point where the record will be almost impossible to beat; as there are natural forces such as gravity and friction that dictate the extent to which humans can jump; as jumping is contingent on one’s muscles take-off force; which also has a natural limit.
There were certain times in the data where there were significant fluctuations, such as the 1904 Olympics, where the record was 10-11 centimeters less than the previous and the post-Olympics (1896 / 1908). This would be considered an outlier in the data. A probable reason for that outlier was the rather low amount of competitors for the 1904 Men’s High Jump; only 6 participants competed. Another factor was that the previous (1900) Olympic champion, Irving Baxter, did not participate in the 1904 event, hence an absence of a competitor who would be likely to score around 190 centimeters.
Another significant fluctuation would be the 1936 to 1948 time period; where the increasing height values decreased in value significantly through that time period. It is most likely that athletes in that time period were not likely to be training competitively for the Olympic level as most countries had a bigger priority of fighting a war; hence funding was likely diverted from athletic pursuits onto military funding; hence the low value of 198 centimeters in 1948 when the Olympics began again.
Finally, there was a significant increase between 1976 and 1980 in the Men’s High Jump; a increase of 11 centimeters. This would be attributed to Wessig’s surprise performance at the 1980 event; Wessig covertly used a new unique technique not used before which contributed to his breaking the world record by a wide margin.
Given the new data expansion, it would seem appropriate to modify the model so that it fits much more with the extra data points. One simple yet highly effective modification would be to take another polynomial regression to all the data points and hence derive a polynomial function that would best serve the new data points. However, doing that will yet not solve the problem when more data is introduced into the equation.
Hence, this time a logarithmic function would be suggested to solve the dilemma of extreme high values; as a logarithmic function would increase less and less as the years move on, which is ideal in this situation and many other competitive record events such as swimming events.
A linear function approach using the Least Squares approach would be time consuming, but provide a simple line with a non-changing gradient that can be beneficial when additional data is exposed; as it is less likely than others to provide extreme values of data. If technology was involved in this; a simple linear regression would also suffice to find out a linear equation that would be “best-fit.”
In conclusion, let us conduct one example of a modification; linear regression - on Autograph to modify our current linear function to improve the best-fit. After inputting data (see Appendix, Figure 5) Let’s conduct linear regression with the programme; which then gives us a best-fit function. Let’s see the next page for a revised linear function.
Graph 10: Revised Linear Function
As is evident from the graph; the revised linear function corresponds with many more points than the original linear function, and in general fits the data points much better than the original function. While improving the linear function still does not address for the constant increase over time; the function works very well in this instance as the gradient of the revised function allows more data points to accurately correspond with the function.
In conclusion, we can find from the data that records do not quickly increase over time, but gradually increase and may sometimes decrease due to extraneous events such as the Second World War. It was also beneficial to realize the natural limits behind real-life examples such as how high a human can naturally jump. This project also demonstrates the limitation in technology; while technology is very accurate within ranges of data that is already given; it cannot accurately predict futures within the graph and certain anomalies such as outliers in the data.
Colophon
The technology used in this assessment includes but is not limited to: TI-84 calculator, AutoGraph Windows Version 3.1, Mac OS Grapher, Windows Publisher, and Microsoft Excel.
Appendix
Figure 1 – Data input onto Autograph
Figure 2 – Calculator display of values
Calculator Screen 1 Calculator Screen 2
Figure 3: Editing data sets on Autograph
Figure 4: Broad view of quintic and linear function
Figure 5: Revissed linear regression data plotting
Black and white: 1-7, 9, 14, 15, 17-19
Colour: Pages 8. 10. 11, 12, 13, 16