Hypothesis - The average word length in Broadsheets will be longer than that in tabloids.

In this investigation I am going to exclude the following things from my counting:

* Names (places, people)

* Headlines

* Acronyms or abbreviated words

* Numbers

* Symbols

Results

Article 1

After counting all the words I then used the standard deviation formula to exclude any extreme results, this will enable me to create more accurate results overall.

The formula for standard deviation is,

I have now created a table so that it is easier for me to calculate the standard deviation of my results

The standard deviation = 4.629734226

This means that all my answers should be within about 4.6 of the mean, i.e. 0 - 9.

4 of my results fall outside this guideline so they will therefore be excluded from my investigation.

I will now repeat the process for my Broadsheet results

5.168194656

This means my data must fall between 5.2 of the mean this is roughly 0 - 10 therefore 2 results will be excluded from my final table where all my data is grouped together.

I have repeated this process 20 times. As I have shown this one example I do not feel it necessary to show the other 19. Instead all the data I collected I have grouped into one big table.

All articles

(The average word length of all the articles in Tabloid and Broadsheet)

From the above information I can now plot a frequency polygon graph, this should give me some idea as to which is the most popular size of word, and how this compares to my average word size.

Frequency Polygon Graph for Tabloid Data

This graph clearly shows that words with three letters are the most the popular. The sharp slope at the beginning shows that the majority of words are between 1 - 3 letters long. The less steep slope after the words with three letters suggests that fewer words have over three letters. This is quite close to mean word length.

Frequency Polygon Graph for Broadsheet Data

This graph clearly shows that words with three letters are the most the popular. The sharp slope at the beginning shows that the majority of words are between 1 - 3 letters long. The less steep slope after the words with three letters suggests that fewer words have over three letters. This is quite close to mean word length.

Frequency Polygon Graph for Tabloid and Broadsheet Data

Because the line for tabloid is above the line for Broadsheet, it suggests that there are much more words of shorter length in a Tabloid newspaper than a Broadsheet newspaper, and vice versa for the latter parts of the graph.

Cumulative Frequency Graph

At the moment I do not have sufficient information to plot a cumulative frequency graph, I must first get my grouped data and create a running total.

Cumulative Frequency Graph for Tabloid and Broadsheet Data

The graph once again shows that there are more shorter words in Tabloid newspapers then there are in Broadsheet newspapers. This goes to prove my original predicition.

Evaluation of Hypothesis 1

After finding the mode and median word lengths for my box and whisker diagram, I have discovered that all the articles have similar word lengths.

I also calculated an accurate average by excluding the extreme values. These results were in agreement with my original prediction, that Broadsheet papers typically use much longer words. This in effect proves the argument that Broadsheet papers are for the 'better educated' people and therefore will be able to use much more advanced vocabulary.

I then drew a frequency polygon and cumulative frequency chart to view the spread of results. When both graphs were put together it clearly showed that the majority of words were of 3 letters long, this . This is reversed at the end, where the higher the mean; the higher the frequency of longer words. The Daily Star has in fact got no words longer than 12 letters long that has passed the standard deviation test. The low numbers of words longer than 12 letters long were found to be extreme values and hence discarded.

The cumulative frequency chart shows that:

* The Daily Star has more shorter words than long - this is shown by the steep gradient in the lower half of the graph and the position at which it becomes a horizontal line.

* There were more extreme values in the Daily Star than any other newspaper-this is shown by the lower finishing point.

* The gentle gradient of the Gardian compared to that of the Daily Star shows that there is a more even spread of results between 2-7 letters long and that also there are considerably greater frequencies of longer words therefore contributing to the higher mean length.

The same features are also shown by the histograms.

The box and whisker diagrams show:

* The Daily Star has a shorter range of word lengths - perhaps giving a possible reason for why the mean length is lower

* The Inter-quartile range of results for the Gardian is greater, - this shows that there is a wide spread of results - accounting for the higher broadsheet word length

* The inter-quartile ranges of results for the Daily Mail and the Yorkshire Evening Post are very similar - perhaps accounting for the similarity of the mean lengths (0.18 letters difference)

The high number of extreme values in the Daily Star can perhaps be explained by the low mean. Due to the low mean, any values which would be acceptable in the Gardian, when standard deviation is used are unacceptable. The size and nature of the sample data can perhaps explain why there is a large spread of results. A larger sample may have given results closer together or given a wider spread of results, 20 articles is sufficient for a reasonable evaluation but not sufficient enough for a firm conclusion to be founded on.