Posts

Showing posts with the label statistics

11.2 Independent samples t-test for a difference of sample means

Image
The Fibobelly data from 10.1 was used to illustrate an independent sample t-test for a difference of sample means.  spreadsheet The line at 1.4992 illustrates a predicted population mean between the two confidence intervals that neither sample is expected to be able to have as a poulation mean. Desmos A Desmos file generates the same p-value using the ttest function. The Desmos file also makes use of being able to generate t-critical and the t-ditribution in the image. 

11.1 t-test for a paired difference in marble masses

Image
Students select two marbles from a box, one from each side of a divided box. One side includes marbles less than or equal to 5.2 grams, the other side contains marbles greater than 5.2 grams.  Data from spring 2026 demonstrated that students could distinguish which marble was more massive using a paired t-test and an alpha of 0.05. The ability of Desmos to display the t-distribution graphically (and calculate t-ciritical from data) allows for the graphical display of the result. The t-test capability displays the p-value based directly on the data. Desmos has become a powerful tool for displaying statistical results graphically. Calculating t-critical involves using tdist(n-1).inversecdf(0.975) for a 95% confidence interval. In the above example the test is an independent samples t-test, not a pairwise test of the differences.  Desmos To do a pairwise t-test the data has to be the pairwise differences, which can also be done as seen above.  Desmos This term ...

Desmos graphs t-distribution and calculates t-critical

Image
Desmos can now calculate t-critical values and graph the t-distribution, an upgrade from only being able to graph the normal distribution. Desmos The normal distribution showed area values larger than 0.95 when upper and lower bounds calculated using the t-distribution were entered. The availability of the t-distribution now shows the correct probability for upper and lower limits calculated using the t-distribution, a necessity when working with sample means. The function for generating t-critical for a sample size n and a 95% confidence interval: The period notation allows one to apply additional functions . These functions allow one to evaluate or manipulate the distribution in different ways. The Desmos file provides a visual approach to confidence interval hypothesis testing calculations done in Google Sheets . The above spreadsheet pulls data from exercises in chapter 9.2 and 10.1 of an MS 150 Statistics course. 

10.1 Confidence interval hypothesis test

Image
  A confidence interval hypothesis test was done on the theory that the navel is located at the golden ratio, that the fibobelly ratio is the golden ratio.  Spreadsheet This was refuted for both the females and the males. The number line approach to interpreting confidence intervals was illustrated on the board. 10.2 and 10.3 then use the data from 9.2 and 10.1 (Paper aircraft flight distance and fibobelly data) in a single combined spreadsheet to illustrate the material in those two sections. The slide deck provides guidance and useful links.

9.2 Paper aircraft confidence interval

Image
This term the aircraft were thrown from the second floor of the east end of the A classroom building. Downstairs after measuring distances, Estele and Tiepo looking out over the field of paper aircraft. Setting up to measure those that went negative. West breeze, very light, El Niño induced. Can you see the plane?  Caught a brief surge in the west wind. Way out near the fence line. Board work Spreadsheet for current term. A multi-term master also exists. 

8.2 Five marbles and the standard error of the mean

Image
Materials required are marbles and a 200 gram scale. As was done last spring , this term a mix of marbles was used.  This term the current term spreadsheet was not properly setup. Column C was accidentally wiped along with columns A and B. This wiped F2, F3, and the corresponding cells below in F.  Students were given five marbles. This avoided an issue that had been created by giving students five of the same marbles. When different sets of five were in use, some were the smaller, lighter blue marbles and these would generally not capture the population mean. A spreadsheet, single tab for the Smartboard memory limitations, is used to generate data.  After the first five marbles were massed the point estimate of the population mean was 5.22. The massing was paused to explain that this was the best estimate of the population mean for that student. The point estimate for the population mean.  By an absolute accident, Kozel wound up with five 5.2 gram marbles, a stand...

7.3 Areas under the normal curve

Image
This lecture was set up by the rolling of dice for sections 7.1 and 7.2 . Data for five dice was extended using the function: =int(rand()*6+1)+int(rand()*6+1)+int(rand()*6+1)+int(rand()*6+1)+int(rand()*6+1)+int(rand()*6+1) The values were then copied and pasted as values only to stop automatic recalculation and freezing the mean and standard deviation. A spreadsheet was built for display in class.  An associated Desmos demonstration file was also added. 

7.1 7.2 2026.3 normal curve introduction with dice groups

Image
This builds off of the concept piloted spring 2026 which used students rolling a single die to generate data columns each of which is a uniform distribution and then the sum is a normal distribution. The result was statistically illustrative but opaque to the students. The whole business of summing arbitrary columns didn't make intuitive sense for a student with no background in probability. Better to have students rolling one die, two dice, three dice, and so forth, entering the sums into columns on the spreadsheet. Then do histograms of these columns. The complication is that those rolling five dice will be vastly slower than those rolling one die. Teams of two are needed. Six groups could be formed. Whether there should be a six die group or a second group of five is yet to be determined. Six dice have 279,936 outcomes and a sum of 36. The students were all added to a Google Sheets spreadsheet as editors. Maireen arrived early and was given five dice. Est...

5.3 50 orange 35 blue 15 white marbles

Image
One hundred marbles. Students blind draw a single marble. Population probability is known. Tally the sample to obtain relative frequency.  Last spring the recommendation was to move to ratio level data . Upon further thought, the ratio level data would not demonstrate the linkage between mathematically determined probabilities and statistically inferred probabilities. Ratio level data wouldn't have obvious population values.  One could use the mass of the marbles to predict a known population mean, but that is a conceptually harder system in which to explain what the probability of individual mass classes ought to be. Those weight class probabilities would be driven by an underlying normal distribution. Students would not comprehend why the predicted probability for the class bounded by the mean minus a standard deviation and the mean has a predicted probability of 34% Fifty orange marbles, thirty-five blue marbles, and fifteen white marbles sets up an easier to se...

4.2 4.3 Slope intercept and slope

Image
Due to 4.2 and 4.3 being combined as the result of a holiday, swizzle data was gathered on the porch to reduce travel time. The period was still tight. A ten meter run on the porch was still functional. Spreadsheet Data was recorded directly into the spreadsheet on the Samsung Active5 tablet while on the porch. Another spreadsheet  tracks 163 runs back to 2011. The board was used to emphasize that the columns were reversed.  Note that the board is using the data from 4.1 in the previous class . This run would be particularly useful because of the large gaps in the split data out beyond 33 meters. That data can be extrapolated beyond maximum x.  The swizzle data is more useful as extrapolation below minimum x or above maximum x are physically unrealistic. Packing the demonstration, 4.2, and 4.3 was tight. Too tight to demonstrate the 4.1 homework and hooping data exploration. 

4.1 4.2 RipStik run

Image
The RipStik run returned to the covered walkway after testing out the new health center fall 2025 and the unfinished student center spring 2026. An 81 meter run from LRC to the A building was laid out. This was a cold open on the sidewalk with a note to students ahead of class to meet on the sidewalk. This saved significant time. Initial roll was called on the sidewalk. O meters was at the LRC.  Every three Meyers was marked out to 33 meters in front of faculty office. 51 meters at the junction. 71 at the turn 81 at the A building. Spreadsheet and Desmos  were prepared in advance. The layout worked well to illustrate the linearity present in the first 33 meters. The gaps to 51m, 71, and 81 meters provided a chance for interpolation. This also was very functional when riding the RipStik. the segment into 51 meters allowed for setting up for the slope into 71 meters. Speed increased to 2.25 meters per secon...